[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6FVEGTM249087JME9HCEWJV":3,"skill-same:01M6FVEGTM249087JME9HCEWJV":77},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":39,"linkOnly":40,"markdownContent":41,"markdownBody":42,"frontMatter":43,"refName":54,"commitSha":55,"skillDirectory":56,"forksCount":57,"repositoryHtmlUrl":58,"repositoryHomepageUrl":59,"ownerHtmlUrl":60,"ownerBlogUrl":59,"ownerDescription":61,"updatedAt":62,"install":63,"resources":76},"01M6FVEGTM249087JME9HCEWJV","vuln-tracker",null,"One lifecycle poll over everything vuln-scanner produces - PR and advisory status, PVR triage transitions, and pending-disclosure aging, with a stars-secured impact headline and one action queue.","aeonfun\u002Faeon","aeonfun","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F287715280?v=4","01M6CZ23C6MXGMF4N0V1DZ1247","aeon","01M6CZ23GMZP1BR424J8Y8K3DG",793,"MIT","https:\u002F\u002Fgithub.com\u002Faeonfun\u002Faeon\u002Fblob\u002Fmain\u002Fskills\u002Fvuln-tracker\u002FSKILL.md",[],[19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38],"vuln tracker","tracker","global","triage","action","github","advisory","aging","gh","dev","lifecycle","stars","pr","security","status","token","write","scanner","impact","Meta","2026-11-10 10:12:30",false,"---\nname: vuln-tracker\ndescription: One lifecycle poll over everything vuln-scanner produces - PR and advisory status, PVR triage transitions, and pending-disclosure aging, with a stars-secured impact headline and one action queue.\nmetadata:\n  title: Vuln Tracker\n  category: dev\n  var: \"\"\n  mode: write\n  tags:\n    - meta\n    - security\n    - github\n  depends_on:\n    - vuln-scanner\n  requires:\n    - GH_TOKEN?\n    - GH_GLOBAL?\n---\n\n> **${var}** — Scope selector for the lifecycle poll:\n> - empty → **full lifecycle poll**: PR\u002Fadvisory status + PVR triage + disclosure-queue aging (default).\n> - `prs` (also `pr` \u002F `tracker`) → **Arm A only** — PR\u002Fadvisory status audit + stars-secured dashboard.\n> - `pvr` → **Arm B only** — PVR triage-state poll on submitted advisories.\n> - `queue` (also `disclosures` \u002F `backlog`) → **Arm C only** — pending-disclosure queue aging + escalation.\n> - a bare `GHSA-xxxx-xxxx-xxxx` → **Arm B, single-advisory mode** — check just that one advisory's triage state on demand.\n\nToday is ${today}. This skill is the daily read\u002Fpoll arm of the vuln pipeline: `vuln-scanner` opens PRs, submits PVRs, and queues disclosure drafts, then moves on. This skill polls everything it produced and surfaces what the operator must look at — it does not open PRs or submit advisories itself (those are `vuln-scanner`'s write actions).\n\n## Voice\n\nIf `soul\u002FSOUL.md` and `soul\u002FSTYLE.md` are populated, read them and match the operator's voice in every notification. If empty or absent, use a clear, direct, neutral tone.\n\n## Capability mode\n\nThis skill runs `mode: write` deliberately. It is a read\u002Fpoll arm, but three of its capabilities cannot run under `read-only`:\n- **Arm B polls private, unpublished advisory triage state** via `gh api repos\u002F$REPO\u002Fsecurity-advisories\u002F$GHSA`. Draft\u002Ftriage advisories are visible only to the repo maintainers and the reporter, so the read is intrinsically authenticated; `read-only` strips `gh`, and a bare `$SECRET` on the command line is refused by the Bash permission layer (so a hand-rolled authenticated `curl` isn't an option either) — the read needs `gh api`, which handles auth internally.\n- **Arm B persists state transitions** — it rewrites `state:`\u002F`last_checked:`\u002F`resolved_at:` frontmatter in `memory\u002Fpending-disclosures\u002F*.md` in place and **moves** resolved files to `memory\u002Fpending-disclosures\u002Fresolved\u002F` (`Edit`\u002F`git mv` — both stripped in `read-only`).\n- **Arm A leans on authenticated `gh api`** for the PVR-state endpoint (`repos\u002F$REPO\u002Fprivate-vulnerability-reporting`) and repo\u002Fadvisory reads.\n\nNo sibling writes a repo file **outside** `memory\u002F`, so there was nothing to relocate — the only outside-`memory\u002F` writes are the ephemeral `.pending-notify-temp\u002F` notify-staging files. Keeping `write` preserves every absorbed capability.\n\n## Shared preamble (run for every invocation)\n\n1. Read `memory\u002FMEMORY.md` for context.\n2. Read the last ~3 days of `memory\u002Flogs\u002F` and drop anything already reported — don't re-surface the same signal twice.\n3. Read `soul\u002FSOUL.md` + `soul\u002FSTYLE.md` if populated (voice).\n4. **Parse `${var}` → scope** (deterministic; trim + lowercase, except a `GHSA-` value which is compared case-insensitively but preserved verbatim):\n   - empty → `scope = full` (run Arm A, then B, then C).\n   - matches `^GHSA-` (case-insensitive) → `scope = pvr`, `single_advisory = \u003Cthe GHSA value>` (Arm B filtered to one advisory).\n   - `prs` \u002F `pr` \u002F `tracker` → `scope = prs` (Arm A only).\n   - `pvr` → `scope = pvr` (Arm B only, all advisories).\n   - `queue` \u002F `disclosures` \u002F `backlog` → `scope = queue` (Arm C only).\n   - anything else → log `VULN_TRACKER_BAD_VAR: unrecognized scope '\u003Cvar>'`, send no notification, exit.\n5. `mkdir -p .pending-notify-temp` and start an empty combined-notification buffer at `.pending-notify-temp\u002Fvuln-tracker-${today}.md`. Each arm that has signal **appends its section** to this buffer; at the very end (step \"Notify\") the skill sends the buffer **once** if it is non-empty. This keeps notifications tight — a full poll with signal in two arms is one message, not two.\n\nThen run the arm(s) selected by `scope`, and finish with the shared **Notify** and **Log** steps.\n\n## Network Note\n\n- **Arm A & Arm B (GitHub reads):** all data via `gh api` \u002F `gh search` \u002F `gh pr view`. `gh` handles auth internally via `GH_TOKEN` (and Arm B's private-advisory reads need the elevated `GH_GLOBAL` PAT). No env-var-authenticated `curl` from bash — a bare `$SECRET` on the command line is refused by the Bash permission layer, so `gh api` (auth handled internally) is the reliable path; no postprocess scripts needed. Arm B keeps a documented `curl` fallback for the advisory endpoint — see Arm B step B2 — but `gh api` is preferred.\n- **Arm C (local only):** reads only local files (`memory\u002Fpending-disclosures\u002F`, `memory\u002Fissues\u002F`, `memory\u002Ftopics\u002Fpr-status.md`). No outbound network or auth required.\n- **Shell shape.** Each Bash call is one allowed command (pipes are fine). The run refuses a call that sets a variable (`BOT_AUTHOR=\"...\"`, `REPO=\"owner\u002Frepo\"`, `PVR_NOW=$(gh api ...)`) or joins commands with `;`, `&&` or `||`, and each refusal costs a model turn. So write the bot author, each repo, PR number and GHSA ID into each `gh` \u002F `.\u002Fsecretcurl` call literally (the `\u003CBOT_AUTHOR>`, `\u003CREPO>`, `\u003CNUM>`, `\u003CGHSA>` placeholders), make one call per PR, repo or advisory instead of a loop, and turn each `|| echo fallback` into a prose rule for when the call fails. Append each arm's section to the notify buffer with Edit (Write when the file is new), never with a shell `>>`.\n\n---\n\n## Arm A — PR & advisory lifecycle status  (scope `full` or `prs`)\n\nAudit the lifecycle status of every disclosure `vuln-scanner` has produced. Without a follow-up loop, three things rot silently:\n\n- **Merged-but-uncelebrated wins** — landed fixes never reach self-improve \u002F reflect (retro) without manual aggregation.\n- **Maintainer questions on open PRs** — a maintainer comments asking for clarification; if the bot doesn't see it, the PR ages out.\n- **Queued drafts past their disclosure window** — entries with `channel: \"skipped\"` (no-safe-channel) vanish into `vuln-scanned.json` with no recurring re-probe.\n\nThis arm cross-references `memory\u002Fvuln-scanned.json` against live GitHub state and surfaces anything the operator should look at.\n\n### A1. Load the canonical scan history\n\n**`memory\u002Fvuln-scanned.json` may be a flat top-level array of scan rows, not a `{scans: [...]}` object** - vuln-scanner writes the flat shape. Read it in a way that accepts both (the flat `[ {repo,scanned_at,findings,channel}, ... ]` shape and the legacy `{scans: [...]}` object), one `jq` call:\n\n```bash\njq -c 'if type==\"array\" then .[] else (.scans[]? \u002F\u002F empty) end' memory\u002Fvuln-scanned.json 2>\u002Fdev\u002Fnull\n```\n\nA bare `jq -c '.scans[]'` **errors** on a flat-array file (`Cannot index array with string \"scans\"`) and, with `2>\u002Fdev\u002Fnull` swallowing the error, silently yields zero rows - so the arm reads an empty scan history and reports only what other passes happen to surface. The `if type==\"array\"` guard is the fix; it keeps the legacy object shape working too.\n\nIf `memory\u002Fvuln-scanned.json` doesn't exist or the parsed row count is 0, log `VULN_TRACKER_SKIP: no scan history` for this arm and skip Arm A (no notification section - first runs of `vuln-scanner` haven't happened yet). In a `full` poll, continue to Arm B\u002FC.\n\nEach scan entry has at minimum: `repo`, `scanned_at`, `findings`, `channel`, `severity`. Public-PR entries also have `pr` (URL). Pending-disclosure entries have `draft_at` and `patch_branch`. Skipped entries have `reason`.\n\n**Retro-active coverage:** if the JSON was written after vuln-scanner started running, some PRs won't be in the JSON. Pull all bot-authored security PRs from GitHub directly (next step) to fill the gap.\n\n### A2. Pull all bot-authored security PRs from GitHub\n\n`vuln-scanner` opens security-fix PRs on a **`security\u002F`** branch. The **title** varies: dependency-bump PRs use **`fix(deps):`** (the A5a template is `fix(deps): bump \u003Cpkg> to patch \u003CCVE>`), and older or hand-written ones use **`fix(security):`** or a bare **`security:`**. **Match all three title prefixes, and treat the `security\u002F` branch prefix as the real invariant** - the branch name is mandated by A5a, the title wording is not.\n\n> **Do not filter on `fix(security):` alone.** vuln-scanner's dominant output is `fix(deps):`, so a `fix(security):`-only filter silently drops most in-flight security PRs - that is how `alibaba\u002Fopen-code-review#541` (conflicted, lint-failing, one unaddressed review) was missed by a poll while other PRs were tracked. Also exclude `docs(security):` - those are `security-sync` website PRs, not disclosures.\n\n`gh search prs --json` does **not** expose `headRefName` - that field is only available via GraphQL. So run **both** passes below every time and union them on `repository + number`; neither alone is complete.\n\nThe bot author is whoever the workflow uses (typically `github-actions[bot]` or a dedicated account configured in the workflow). Determine the author from `aeon.yml` or the workflow file; default to whatever account opened the most recent security-fix PR (any of the prefixes above) you can find.\n\n**Pass 1 (title prefixes, works everywhere)**, with the resolved bot author written in literally as `\u003CBOT_AUTHOR>`:\n\n```bash\ngh search prs --author \"\u003CBOT_AUTHOR>\" --json number,title,url,state,createdAt,closedAt,repository --limit 200 \\\n  | jq '[.[] | select(\n      (.title | startswith(\"fix(deps):\")) or\n      (.title | startswith(\"fix(security):\")) or\n      (.title | startswith(\"security:\"))\n    )]'\n```\n\n**Pass 2 (GraphQL, required - picks up every `security\u002F` branch whatever the title says)**, same author written in literally:\n\n```bash\ngh api graphql -f query='\n{\n  search(query: \"author:\u003CBOT_AUTHOR> is:pr sort:created-desc\", type: ISSUE, first: 100) {\n    nodes { ... on PullRequest {\n      number title url state createdAt closedAt mergedAt\n      repository { nameWithOwner }\n      headRefName\n    }}\n  }\n}' | jq '[.data.search.nodes[]\n    | select((.headRefName \u002F\u002F \"\") | startswith(\"security\u002F\"))]'\n```\n\nUnion the two result sets, dedup by `repository + number` (equivalently, by URL).\n\nCross-reference with `vuln-scanned.json`:\n- PR present in JSON → use JSON's `severity` \u002F `cwe` \u002F `note` for the row.\n- PR not in JSON → mark as `pre-history`; fill severity from the PR title if obvious.\n- JSON entry with `channel != \"public-pr\"` → no PR to fetch; goes in the \"queued\" \u002F \"skipped\" sections.\n\n### A3. Fetch live state for each open PR\n\nFor each open PR (state from step A2), one call per PR, with `\u003CREPO>` and `\u003CNUM>` written in literally:\n\n```bash\ngh pr view \"\u003CNUM>\" -R \"\u003CREPO>\" --json state,merged,closedAt,createdAt,reviews,comments,reviewDecision,author\n```\n\nPer-PR signals:\n- **Maintainer-needs-answer**: any comment whose `author.login != $BOT_AUTHOR` posted **after** the most recent comment by `$BOT_AUTHOR` (or after PR creation if the bot hasn't commented). Also `reviewDecision == \"CHANGES_REQUESTED\"` always counts.\n- **Stale-no-review**: `state == \"OPEN\"` AND no review AND no maintainer comment AND `createdAt` > 7d ago.\n- **Aging-with-engagement**: `state == \"OPEN\"` AND any maintainer activity AND open > 14d.\n\nIf a scan entry has `advisory_ids` (one or more GHSA IDs), check each one's published state, one call per GHSA ID with the origin repo and ID written in literally:\n\n```bash\ngh api \"repos\u002F\u003CORIGIN_REPO>\u002Fsecurity-advisories\u002F\u003CGHSA_ID>\" --jq '.state \u002F\u002F \"not found\"'\n```\n\n`state == \"published\"` → public advisory visible. 404 → osv-scanner referenced it but the upstream repo never published its own advisory.\n\n### A4. Fetch star counts for every secured repo\n\nFor every unique `repo` across the union from step A2 (JSON history + bot-authored security PRs), one call per repo with the repo written in literally:\n\n```bash\ngh api \"repos\u002F\u003CREPO>\" --jq '{stars: .stargazers_count, archived: .archived}' 2>\u002Fdev\u002Fnull\n```\n\n**Refetch every run.** Do NOT carry star counts forward from the previous `memory\u002Ftopics\u002Fvuln-followup.md` — per-repo counts drift between runs and the secured-stars headline is the operator's load-bearing metric. Cache only within a single run, keyed by `nameWithOwner`, so a repo with multiple PRs is fetched once.\n\nRepo-state handling:\n- **200 with stars**: use `.stargazers_count` (raw integer).\n- **200 with `archived: true`**: still use the star count, but suffix the repo cell with ` (archived)` so the operator knows the maintainer isn't responsive.\n- **404 \u002F 403**: repo was deleted, renamed, or made private. Record `null` and render as `repo-deleted`. Exclude from `total_stars_*` aggregates entirely so dead repos don't quietly zero out the totals.\n- **Other non-2xx**: record `null` and render `★?`. Flag in the run log for operator follow-up.\n\nThese per-repo counts power both the **Stars Secured** aggregate (step A6) and the `Stars` column on every per-repo table in step A6.\n\n### A5. Re-probe `channel: \"skipped\"` and `channel: \"pending-disclosure\"` repos\n\nFor each historical entry where the disclosure couldn't ship, re-check whether the situation changed:\n\n- **`channel: \"skipped\"`** with `reason` containing \"no PVR\":\n  ```bash\n  gh api \"repos\u002F\u003CREPO>\u002Fprivate-vulnerability-reporting\" --jq .enabled\n  ```\n  It prints PVR_NOW. If the call fails, PVR_NOW is `false`. If `PVR_NOW=true` and the original was `false`, surface as **newly-actionable**.\n- **`channel: \"skipped\"`** with `reason` containing \"no SECURITY.md\" - re-check `gh api repos\u002F\u003CREPO>\u002Fcontents\u002FSECURITY.md` and `.github\u002FSECURITY.md`. If now present, surface as **newly-actionable**.\n- **`channel: \"pending-disclosure\"`** — cross-reference with `memory\u002Fpending-disclosures\u002F` to see if the draft is still on disk. If the file is gone but the JSON entry says `pending-disclosure`, mark as `lost-draft` so it stops being escalated forever.\n\n### A6. Categorize every entry, then rewrite the dashboard\n\n| Status | Meaning |\n|---|---|\n| `merged` | PR merged. One-time celebration — drop from notifications after 30d. |\n| `open-clean` | PR open, no maintainer activity yet, \u003C 7d old. Wait. |\n| `needs-answer` | Maintainer commented or requested changes. **Operator action.** |\n| `stale-no-review` | Open > 7d, zero maintainer activity. Consider polite ping or close. |\n| `aging-engaged` | Open > 14d with engagement. Operator should triage. |\n| `closed-no-merge` | PR closed without merging. Capture the reason for review. |\n| `queued` | `pending-disclosure` draft on disk, not yet shipped. |\n| `skipped-rechecked` | Channel was \"skipped\" originally; re-probe still shows no channel. |\n| `newly-actionable` | Skipped originally; PVR or SECURITY.md now present. **Operator action.** |\n| `lost-draft` | JSON says pending-disclosure but draft file is gone. Display once, then suppress. |\n| `pre-history` | PR found via search but predates `vuln-scanned.json`. Fill what we can. |\n\nThen rewrite `memory\u002Ftopics\u002Fvuln-followup.md` (rewrite — don't append; this file is a living dashboard, not a log).\n\nThe **Stars Secured** block goes at the top so the operator sees aggregate impact before drilling into rows. `total_stars_secured` = sum of stargazers across every unique repo where vuln-scanner has landed at least one merged PR. `total_stars_in_flight` = sum across repos with an open PR. `total_stars_tracked` = sum across the full union. Track all three because celebration uses `secured`, prioritization uses `in_flight`, and historical review uses `tracked`.\n\nRound star counts to abbreviated form for the headline (12.4k, 1.8k, 940). Keep raw integers in the per-repo tables so sort\u002Fdiff stays exact.\n\n```markdown\n# Vuln Tracker Status\n\n*Last updated: ${today}*\n\n## Stars Secured\n\n- **Merged-PR repos (secured):** ★ \u003Ctotal_stars_secured> across \u003Csecured_repo_count> repos\n- **Open-PR repos (in flight):** ★ \u003Ctotal_stars_in_flight> across \u003Cin_flight_repo_count> repos\n- **All tracked repos:** ★ \u003Ctotal_stars_tracked> across \u003Ctotal_repo_count> repos\n\n### Secured leaderboard - every merged PR ranked by repo stars\n| Rank | Repo | Stars | PR | Merged | Severity | Title |\n|------|------|-------|----|--------|----------|-------|\n\n### Per-repo breakdown - secured (sorted by stars desc)\n| Repo | Stars | Merged PRs | First merge | Latest merge | Severities landed |\n|------|-------|------------|-------------|--------------|-------------------|\n\n### Per-repo breakdown - in flight (sorted by stars desc)\n| Repo | Stars | Open PRs | Oldest open | Severities open |\n|------|-------|----------|-------------|-----------------|\n\n### Per-repo breakdown - queued \u002F skipped \u002F closed (sorted by stars desc)\n| Repo | Stars | Status | Severity | Note |\n|------|-------|--------|----------|------|\n\n## Operator-action queue\n\n### Needs answer (\u003Ccount>)\n| Repo | Stars | PR | Title | Last activity | Latest commenter |\n|------|-------|----|----|--------------|------------------|\n\n### Newly actionable - channel opened up since the original scan (\u003Ccount>)\n| Repo | Stars | Original date | Original blocker | Now |\n|------|-------|---------------|------------------|-----|\n\n### Stale or aging\n| Repo | Stars | PR | Age | Status | Suggested action |\n|------|-------|----|----|-----|------------------|\n\n## Recently merged (last 30d, \u003Ccount>)\n| Date merged | Repo | Stars | PR | Severity | Title |\n|-------------|------|-------|----|----------|-------|\n\n## Open \u002F clean (no operator action - wait, \u003C 7d) (\u003Ccount>)\n| Repo | Stars | PR | Severity | Opened | Age |\n|------|-------|----|----------|--------|-----|\n\n## Closed without merge (last 30d, \u003Ccount>)\n| Date | Repo | Stars | PR | Severity | Title | Likely reason |\n|------|------|-------|----|----------|-------|---------------|\n\n## Queued (no PR yet) (\u003Ccount>)\n| Severity | Repo | Stars | Original channel | Original blocker | Days queued |\n|----------|------|-------|------------------|------------------|-------------|\n\n## Lost-draft ghosts (suppressed from notifications)\n| Date | Repo | Stars | Severity |\n```\n\n### A7. Decide whether Arm A has signal\n\n**No signal (contribute nothing to the notification)** if all are true:\n- Zero `needs-answer`\n- Zero `newly-actionable`\n- Zero items moved to `merged` since the last run\n- Zero items moved to `closed-no-merge` since the last run\n- Zero items aged into `stale-no-review` or `aging-engaged` since the last run\n\nTo detect \"since last run,\" diff today's categorization against the previous `memory\u002Ftopics\u002Fvuln-followup.md`. If the file doesn't exist (first run), treat all entries as new and surface the full backlog.\n\n### A8. Emit Arm A's notification section (when it has signal)\n\nAppend this section to the combined buffer (`.pending-notify-temp\u002Fvuln-tracker-${today}.md`). Keep the whole message under 4000 chars; if tight, drop the merged\u002Fopened\u002Fstale section bodies and keep counts only.\n\n```\n*PR & advisory status*\n\n★ secured: \u003Ctotal_stars_secured> across \u003Csecured_repo_count> repos  (in flight: \u003Ctotal_stars_in_flight> \u002F \u003Cin_flight_repo_count>)\n\nneeds answer: \u003CN>\n\u003Crepo> (★\u003Cstars>) #\u003Cnum> - \u003Clatest_commenter>: \"\u003Ccomment_excerpt_first_120_chars>\"\n\nnewly actionable: \u003CN>\n\u003Crepo> (★\u003Cstars>, queued \u003Cdays>d) - \u003Cwhat_changed>\n\nmerged this week: \u003CN>  \u003Clist_with_stars>\nopened, waiting: \u003CN>\nstale: \u003CN>\nqueued: \u003CN> (\u003Ccritical>C \u002F \u003Chigh>H \u002F \u003Cother>M+L)\n\nleaderboard top-3 (PRs by ★): #1 \u003Crepo1> ★\u003Cs1> (PR #\u003Cpr1>) - #2 \u003Crepo2> ★\u003Cs2> (PR #\u003Cpr2>) - #3 \u003Crepo3> ★\u003Cs3> (PR #\u003Cpr3>)\n\ndashboard: memory\u002Ftopics\u002Fvuln-followup.md (full leaderboard inside)\n```\n\nRecord Arm A status `VULN_TRACKER_OK` for the log.\n\n---\n\n## Arm B — PVR triage state  (scope `full`, `pvr`, or a bare `GHSA-…`)\n\n`vuln-scanner`'s re-submit arm (`var=resubmit`, the former `pvr-watchlist`) monitors repos *waiting to open* PVR. This arm monitors PVRs that have **already been submitted** and tracks their lifecycle: `triage` → `draft` (accepted) → `published` (public) or `withdrawn` (rejected). Without this, submitted advisories sit unmonitored until manually recalled from memory.\n\nSource of truth: `memory\u002Fpending-disclosures\u002F*.md` files with `channel: pvr` frontmatter. Each such file must have `ghsa`, `repo`, `state`, `submitted_at` fields (full schema below).\n\n**Configuration — optional tracking issue.** The arm can reference an optional **tracking issue** in the operator's own repo — useful for cross-linking advisory state with an internal issue board. Resolve from (priority order):\n1. `aeon.yml` top-level key `pvr_triage.tracking_issue:` (e.g. `pvr_triage: { tracking_issue: \"owner\u002Frepo#123\" }`)\n2. environment variable `AEON_PVR_TRACKING_ISSUE`\n3. unset → skip cross-linking entirely.\n\nIf a tracking issue is configured, mention its URL in the notification and the per-advisory write-up so the operator can navigate to the canonical tracker.\n\n### B1. Discover in-flight PVRs\n\nScan `memory\u002Fpending-disclosures\u002F` for all `.md` files. Parse the YAML frontmatter. Keep only those with `channel: pvr`.\n\nIf `single_advisory` is set (a bare `GHSA-…` was passed), filter to just the matching `ghsa` value (one-off mode).\n\nIf no PVR files found:\n```\nPVRT_SKIP: no submitted PVRs on disk\n```\nRecord that Arm B status and skip Arm B (no notification section). In a `full` poll, continue to Arm C.\n\n### B2. Probe each advisory's triage state\n\nFor each entry, determine `repo` and `ghsa` from frontmatter. Then one call per entry, with them written in literally (for example `owner\u002Frepo` and `GHSA-xxxx-xxxx-xxxx`):\n\n```bash\ngh api \"repos\u002F\u003CREPO>\u002Fsecurity-advisories\u002F\u003CGHSA>\" \\\n  --jq '{state: .state, cve_id: .cve_id, published_at: .published_at}' 2>&1\n```\n\nExpected outcomes:\n\n| Response | Meaning |\n|----------|---------|\n| `{state: \"triage\", ...}` | Maintainer hasn't reviewed yet |\n| `{state: \"draft\", ...}` | Accepted — maintainer is working on it |\n| `{state: \"published\", ...}` | Published — fully resolved |\n| `{state: \"withdrawn\", ...}` | Rejected or withdrawn by reporter |\n| HTTP 403 | Private advisory, we don't have read access — state unknown, treat as still `triage` |\n| HTTP 404 | Advisory deleted \u002F repo private \u002F GHSA invalid — flag as `not-found` |\n\n**Fallback:** `gh api` uses `GH_TOKEN` internally (workflow wires `GH_GLOBAL` for the elevated advisory read). If `gh` is somehow unavailable, fall back to `.\u002Fsecretcurl` with the `{GH_TOKEN}` placeholder (same token; `GH_GLOBAL` itself is not a substitutable name):\n```bash\n.\u002Fsecretcurl -s -H \"Authorization: Bearer {GH_TOKEN}\" \\\n  \"https:\u002F\u002Fapi.github.com\u002Frepos\u002F\u003CREPO>\u002Fsecurity-advisories\u002F\u003CGHSA>\" \\\n  | grep -o '\"state\":\"[a-z]*\"'\n```\n(Never write a raw `curl` with `$GH_GLOBAL` \u002F `$GH_TOKEN`: a bare `$SECRET` on the command line is refused by the Bash permission layer. `gh api` stays the primary path.)\n\n### B3. Detect state changes\n\nCompare the probed `state` to the `state` in the frontmatter.\n\n- **No change:** note it, continue.\n- **Changed:** this is the primary event. Log old → new state.\n\nAlso flag:\n- **Aged triage:** `state=triage` AND (`today` − `submitted_at`) > 30 days → escalate. Most maintainers respond within 30 days; silence past that is actionable.\n- **Accepted (draft):** surface the patch branch from the `patch_branch` frontmatter field — maintainer may want a PR instead of a private advisory.\n- **Published:** advisory is live. The finding is closed. Update state and mark for removal.\n- **Withdrawn:** rejected. Note the reason if visible. Mark for cleanup.\n\n### B4. Update frontmatter in-place\n\nFor each file with a state change, rewrite just the `state` field in the YAML frontmatter. Also update a `last_checked` field (add it if absent).\n\nDo NOT modify the body of the advisory file — only update frontmatter.\n\nExample frontmatter update:\n```yaml\nstate: draft          # was: triage\nlast_checked: 2026-05-21\n```\n\nFor `published` or `withdrawn` entries, add:\n```yaml\nresolved_at: 2026-05-21\n```\n\n### B5. Decide whether Arm B has signal\n\n- **All entries still `triage`, no changes, none aged:** no notification section. Record status and continue.\n- **Any state change, aged entry, or action item:** emit the section below.\n\n### B6. Emit Arm B's notification section (when it has signal)\n\nAppend to the combined buffer:\n\n```\n*PVR triage*\n\npvr triage: {total} advisories in flight. {changed_count} changed.\n\nCHANGED:\n- {repo} {ghsa} - {old_state} → {new_state}\n  {action_item}\n\nAGED (>30d no response):\n- {repo} {ghsa} - {days}d in triage. {severity}. escalate or close.\n  patch: {patch_branch}\n\nSTILL TRIAGE:\n{n} advisories waiting. oldest: {repo} ({days}d).\n\n{if tracking_issue configured}\ntracker: {tracking_issue_url}\n{end}\n```\n\nAction items by transition:\n- `triage → draft` → \"maintainer accepted — offer to PR the patch branch: {patch_branch}\"\n- `triage → published` → \"published as {cve_id}. remove from tracking.\"\n- `triage → withdrawn` → \"rejected. remove from tracking and note in vuln-scanned.json.\"\n- aged triage (>30d) → \"30d+ no response. consider pinging maintainer or withdrawing.\"\n\n### B7. Clean up resolved entries\n\nFor entries where `state=published` or `state=withdrawn` AND `resolved_at` is set: move the file from `memory\u002Fpending-disclosures\u002F` to `memory\u002Fpending-disclosures\u002Fresolved\u002F` (create the directory if needed; use `git mv`).\n\nDo NOT delete — keep as a historical record.\n\nRecord Arm B status `PVRT_OK` for the log.\n\n---\n\n## Arm C — Disclosure-queue aging  (scope `full` or `queue`)\n\nMonitor the pending vulnerability disclosure backlog. `vuln-scanner` queues draft advisories to `memory\u002Fpending-disclosures\u002F` when PVR auto-submission fails or the disclosure path is email-only. Without daily visibility, CRITICAL\u002FHIGH advisories silently age past responsible-disclosure windows. This arm surfaces the queue state and escalates when findings sit too long. (Arm B tracks the lifecycle of the `channel: pvr` subset; Arm C ages the **whole** queue by severity and disclosure status.)\n\n### C1. Scan the backlog\n\n```bash\nls memory\u002Fpending-disclosures\u002F 2>\u002Fdev\u002Fnull\n```\n\nIf the directory doesn't exist or is empty: log `DISCLOSURE_TRACKER_SKIP: no pending advisories` for this arm and skip Arm C (no notification section).\n\n### C2. Parse each advisory file\n\nFor each `.md` file in `memory\u002Fpending-disclosures\u002F` (ignore the `resolved\u002F` subdir):\n\n**From the filename** (pattern: `{repo-slug}-{YYYY-MM-DD}.md` or `{repo-slug}-{YYYY-MM-DD}-{ampm}.md`):\n- Extract target repo slug (everything before the last date segment)\n- Extract filed date\n\n**From the YAML frontmatter** (if present) parse:\n- `repo:` — overrides the filename slug when present (canonical target)\n- `severity:` — CRITICAL \u002F HIGH \u002F MEDIUM \u002F LOW\n- `status:` — see step C2.5 for the controlled vocabulary\n\n**From the file content** (fallback for files without frontmatter), look for these near the top of the file:\n- `Severity:` or `**Severity:**` — one of CRITICAL \u002F HIGH \u002F MEDIUM \u002F LOW\n- `CVE\u002FCWE:` or similar identifier\n- Short title (first non-blank heading line)\n\nIf severity is not parseable, treat as MEDIUM. Compute age: `today - filed date` in days.\n\n### C2.5. Classify each advisory's disclosure state\n\nBefore counting any draft as a past-threshold escalation, decide whether the draft is genuinely pending or already covered. A draft can be in one of these states:\n\n- `escalate` — pending, no canonical PR found, past the severity-tier threshold\n- `pending` — pending, no canonical PR found, within the threshold window\n- `operator-todo` — needs operator-only action (email send, PVR enable nudge); not an agent failure\n- `covered-by-pr` — a canonical disclosure PR has already been filed against the target repo and is `OPEN` or recently merged\n- `superseded-upstream` — the bypass \u002F vuln is fixed in upstream already, draft is dead-weight\n- `submitted` — already submitted via PVR \u002F GHSA; awaiting maintainer response\n\nResolution rules:\n\n1. **Check frontmatter `status:` first** — map the literal value to a state:\n   - `superseded-upstream` → `superseded-upstream`\n   - `submitted`, `submitted-via-pvr`, `disclosed-via-pr-{N}`, `email-sent` → `submitted` or `covered-by-pr`\n   - `pending-operator-send` **with `auto_send: true`** → `pending` (`vuln-scanner`'s disclose arm, `var=disclose`, sends it autonomously - not a human task)\n   - `pending-operator-send` (without `auto_send: true`), `queued for operator manual send`, `email-failed`, any string mentioning \"operator\" → `operator-todo`\n   - `pending`, blank, or missing → fall through to rule 2\n\n2. **Cross-reference `memory\u002Ftopics\u002Fpr-status.md`** (if present) — grep for the `{repo}` slug (frontmatter `repo:` or filename) in the Open section and Recent Merges section. If a row exists with a `fix(security)` or `chore(security)` title against that repo, opened on or after the draft's `detected_at` \u002F `reconstructed_at` \u002F filed-date, classify as `covered-by-pr` and capture the PR number \u002F title for the summary. If `memory\u002Ftopics\u002Fpr-status.md` doesn't exist, skip this lookup and fall to rule 3.\n\n3. **Fall through** — if no status hint and no canonical PR found, classify as `pending`. Then check age vs the severity-tier threshold (CRITICAL 3d \u002F HIGH 7d \u002F MED-LOW 14d) — if past, promote to `escalate`.\n\nThis is the load-bearing step. Without cross-referencing already-merged fix PRs the tracker generates false-positive escalations for drafts that have already been resolved.\n\n### C3. Build the summary\n\nGroup advisories by **state first**, then by severity within each state. The three buckets are:\n\n- **Escalate** — `escalate` state only (truly stuck, past threshold, no canonical PR)\n- **Operator-todo** — `operator-todo` state (email-only sends, PVR-enable nudges, anything awaiting human action)\n- **Cleanup candidates** — `covered-by-pr`, `submitted`, `superseded-upstream` (draft files that can be removed from `memory\u002Fpending-disclosures\u002F`)\n\nSeverity tiers and thresholds (only apply to `escalate` and `pending` states):\n- **CRITICAL** (age threshold: 3 days — escalate immediately)\n- **HIGH** (age threshold: 7 days — escalate at 7d)\n- **MEDIUM \u002F LOW** (threshold: 14 days)\n\nFor each advisory, produce one line:\n```\n- {repo-slug} | {severity} | {age}d | {short title}{state-suffix}\n```\nwhere `{state-suffix}` is empty for `escalate` \u002F `pending`, `[operator-todo: {reason}]` for operator-todo, and `[covered: PR #{N}]` \u002F `[superseded-upstream]` \u002F `[submitted]` for cleanup candidates.\n\nCount totals. Identify advisories in the `escalate` state — those are the only ones that drive the urgent notification path.\n\n### C4. Check for upstream PVR \u002F token issues\n\nLook in `memory\u002Fissues\u002FINDEX.md` for any open issues tagged with `pvr`, `repository_advisories`, or `missing-secret` that explain why advisories are stuck. If such an issue exists, note:\n- Number of consecutive PVR failures (if logged)\n- Fix estimate from the issue notes\n- How many of the backlogged advisories are blocked by it\n\nIf no such issue exists, treat the queue as routine and skip the \"blocked by\" line in the notification.\n\n### C5. Decide whether Arm C has signal\n\nCompute counts from step C3:\n- `escalate_count` — drafts in the `escalate` state\n- `pending_count` — drafts in `pending` (in-window) state\n- `operator_todo_count` — drafts awaiting operator action\n- `cleanup_count` — drafts in `covered-by-pr` \u002F `submitted` \u002F `superseded-upstream`\n\nDecision:\n- **Queue empty**: log `DISCLOSURE_TRACKER_SKIP: queue empty`, no section.\n- **`escalate_count` > 0**: emit the **urgent** section.\n- **`escalate_count` == 0 but `cleanup_count` > 0**: emit the **daily digest** section (includes the cleanup-candidate list so the operator can prune `memory\u002Fpending-disclosures\u002F`).\n- **All `pending` \u002F `operator-todo`, nothing past threshold, no cleanup candidates**: emit the **daily digest** section.\n\nCoverage from `covered-by-pr` \u002F `submitted` \u002F `superseded-upstream` is **never** counted as an escalation — informational only. Operator-todo is **never** an escalation either; it's surfaced separately so the operator knows their inbox.\n\n### C6. Emit Arm C's notification section\n\nAppend to the combined buffer.\n\n**Urgent format** (`escalate_count` > 0):\n\n```\n*Disclosure queue*\n\ndisclosure queue: {escalate_count} past threshold (of {total} drafts).\n\nESCALATE:\n- {repo} - {severity}, {age}d old (threshold: {N}d)\n[... others in `escalate` state ...]\n\noperator-todo ({operator_todo_count}):\n- {repo} - {severity}, {age}d - {operator-reason}\n\ncleanup candidates ({cleanup_count}):\n- {repo} - [covered: PR #{N} \u002F superseded-upstream \u002F submitted] - safe to delete from memory\u002Fpending-disclosures\u002F\n\n{IF blocking issue tracked in memory\u002Fissues\u002FINDEX.md}\nblocked by {ISS-ID} - {short reason}\nfix: {fix estimate} unblocks {N} of {escalate_count + pending_count}\n{end}\n```\n\n**Daily digest format** (no escalation):\n\n```\n*Disclosure queue*\n\ndisclosure queue: {total} drafts. {critical_count} CRITICAL, {high_count} HIGH, {other_count} MED\u002FLOW.\n{pending_count} in-window, {operator_todo_count} operator-todo, {cleanup_count} cleanup candidates.\noldest in-window: {repo} ({age}d).\n{cleanup section if cleanup_count > 0}\n{IF blocking issue tracked in memory\u002Fissues\u002FINDEX.md}\nblocked by {ISS-ID} - {short reason}.\n{end}\n```\n\nRecord Arm C status `DISCLOSURE_TRACKER_OK` for the log.\n\n---\n\n## Notify (shared, runs once at the end)\n\nIf the combined buffer `.pending-notify-temp\u002Fvuln-tracker-${today}.md` has at least one arm section, prepend the header line `*Vuln Tracker — ${today}*` and send once:\n\n```bash\n.\u002Fnotify -f .pending-notify-temp\u002Fvuln-tracker-${today}.md\n```\n\nIf the buffer is empty (every arm that ran was silent), **send nothing** — a clean poll is silent. Keep the whole message under 4000 chars; if over, trim Arm A's body first (keep counts), then Arm C's cleanup list, keeping every `needs-answer` \u002F `newly-actionable` \u002F `ESCALATE` line.\n\n## Log (shared, runs once at the end)\n\nAppend to `memory\u002Flogs\u002F${today}.md` under ONE heading (the health loop parses `### \u003Cskill-name>`), with a discriminator line naming the scope\u002Farms that ran:\n\n```\n### vuln-tracker\n- scope: \u003Cfull | prs | pvr | queue | single-advisory:GHSA-…>\n- Arm A (PR\u002Fadvisory): \u003Cstatus> - \u003Ctotal_in_json> JSON + \u003Cpre_history_prs> pre-history = \u003Ctotal> tracked; states: \u003Cmerged> merged \u002F \u003Copen_clean> open-clean \u002F \u003Cneeds_answer> needs-answer \u002F \u003Cstale> stale \u002F \u003Caging> aging \u002F \u003Cclosed> closed \u002F \u003Cqueued> queued \u002F \u003Cskipped> skipped-rechecked \u002F \u003Cnewly_actionable> newly-actionable\n- Stars: ★\u003Ctotal_stars_secured> secured (\u003Csecured_repo_count> repos) \u002F ★\u003Ctotal_stars_in_flight> in flight (\u003Cin_flight_repo_count> repos) \u002F ★\u003Ctotal_stars_tracked> tracked (\u003Ctotal_repo_count> repos)\n- Arm B (PVR triage): \u003Cstatus> - checked \u003Ctotal>; changed \u003Cchanged_count> (\u003Clist>); aged>30d \u003Caged_count>; still-triage \u003Cwaiting_count>; tracking issue \u003Curl|none>\n- Arm C (disclosure queue): \u003Cstatus> - \u003Ctotal> drafts (\u003Ccritical_count> CRITICAL \u002F \u003Chigh_count> HIGH \u002F \u003Cother_count> MED\u002FLOW); \u003Cescalate_count> escalate \u002F \u003Cpending_count> in-window \u002F \u003Coperator_todo_count> operator-todo \u002F \u003Ccleanup_count> cleanup; oldest in-window \u003Crepo> (\u003Cage>d); blocking issue \u003CISS-ID|none>\n- Operator queue this run: \u003Cneeds_answer + newly_actionable + aged PVRs + escalate_count>\n- Notification: \u003Csent | skipped (no movement)>\n- VULN_TRACKER_OK\n```\n\nOnly include the Arm B \u002F Arm C lines for arms that actually ran under the current scope. Use each arm's own status token (`VULN_TRACKER_OK` \u002F `VULN_TRACKER_SKIP`, `PVRT_OK` \u002F `PVRT_SKIP`, `DISCLOSURE_TRACKER_OK` \u002F `DISCLOSURE_TRACKER_SKIP`) inline, and close with `VULN_TRACKER_OK` for the hub run (or `VULN_TRACKER_BAD_VAR` if the var was unrecognized).\n\n## Required Env Vars\n\n- `GH_TOKEN` \u002F `GITHUB_TOKEN` — GitHub read token for Arm A's PR\u002Fcomment\u002Frepo\u002Fadvisory reads and the PVR-state endpoint. `repo` scope is enough. Present by default in GitHub Actions.\n- `GH_GLOBAL` — GitHub PAT with `public_repo` + `repository_advisories:write` scope, used by Arm B to read the triage state of **private\u002Funpublished** advisories you submitted. Same token `vuln-scanner` uses (all arms). Without it, Arm B advisory reads may 403 (treated as still `triage`).\n- `AEON_PVR_TRACKING_ISSUE` (optional) — internal tracking-issue reference for Arm B cross-linking (else resolved from `aeon.yml` `pvr_triage.tracking_issue:`, else skipped).\n- Arm C requires no env vars — all its data comes from local files written by `vuln-scanner`.\n\n## Pending-disclosure file schema (Arm B & Arm C)\n\n`memory\u002Fpending-disclosures\u002F*.md` files. Newer drafts use YAML frontmatter:\n\n```yaml\n---\nrepo: owner\u002Frepo\nghsa: GHSA-xxxx-xxxx-xxxx\nghsa_url: https:\u002F\u002Fgithub.com\u002Fowner\u002Frepo\u002Fsecurity\u002Fadvisories\u002FGHSA-xxxx-xxxx-xxxx\nchannel: pvr                       # Arm B tracks only channel: pvr entries\nstate: triage                      # triage | draft | published | withdrawn\nstatus: pending-operator-send      # optional; Arm C vocabulary below\nsubmitted_at: 2026-05-12T19:54:42Z\nlast_checked: 2026-05-15           # added\u002Fupdated by Arm B\nseverity: HIGH\ncwe: [CWE-xxx]                      # may be a string or an array — handle both\npatch_branch: https:\u002F\u002Fgithub.com\u002F\u003Cfork-owner>\u002Frepo\u002Ftree\u002Fsecurity\u002Fbranch-name\npatch_commit: abc1234\nsubmit_url: https:\u002F\u002Fgithub.com\u002Fowner\u002Frepo\u002Fsecurity\u002Fadvisories\u002Fnew\n---\n\n# {Repo}: {Title}\n...\n```\n\n- **Arm B required fields:** `repo`, `ghsa`, `channel: pvr`, `state`, `submitted_at`. Optional: `patch_branch`, `patch_commit`, `cwe`, `ghsa_url`.\n- **Older drafts** use inline `**Severity:**` lines and no frontmatter — parse defensively: grep for `severity:` and `Severity:` case-insensitively; if unparseable, default to MEDIUM.\n\n### `status:` controlled vocabulary (Arm C, step C2.5)\n\nSet by `vuln-scanner` \u002F operator \u002F cleanup chores:\n\n- (blank or missing) — pending; falls through to the PR-status cross-ref\n- `pending` — same as blank\n- `pending-operator-send` **with `auto_send: true`** - queued for autonomous send by `vuln-scanner`'s disclose arm (`var=disclose`); classify as `pending` (NOT operator-todo). It flips to `email-sent` on the next disclose run, which only happens when `vuln-scanner` is dispatched with `var=disclose` (the default schedule runs the scan arm). If it lingers armed-but-unsent across several days, either no disclose run was dispatched or `RESEND_API_KEY` isn't configured - surface it once so the operator schedules or runs the disclose arm and wires up Resend.\n- `pending-operator-send` **with `auto_send: false`\u002Fabsent** \u002F `queued for operator manual send` — operator-todo (human must send; e.g. non-email contact or AI-report ban)\n- `email-sent` - sent by `vuln-scanner`'s disclose arm via Resend; awaiting maintainer reply. Treat like `submitted` (informational cleanup candidate, **never** an escalation).\n- `email-failed` - the disclose arm gave up after repeated send failures (bad address \u002F Resend error). **operator-todo** — the contact needs fixing or a manual send.\n- `submitted` \u002F `submitted-via-pvr` — submitted, awaiting maintainer\n- `disclosed-via-pr-\u003CN>` — covered-by-pr, draft can be archived\n- `superseded-upstream` — bypass is already fixed in upstream; draft is dead\n- Any string containing `operator` — operator-todo\n\nWhen a canonical PR lands but the draft's `status:` was never set, Arm C falls through to step C2.5 rule 2 (cross-ref `pr-status.md`) and classifies as `covered-by-pr` automatically. The `status:` shortcut just bypasses the grep.\n\n**When to delete a draft:** cleanup candidates in the notification can be removed by the operator with `rm memory\u002Fpending-disclosures\u002F\u003Cfile>.md`. Safe once the canonical PR is open — the patch branch on the fork remains the authoritative artifact.\n\n## Notes & related\n\n- **`vuln-scanned.json` schema is loose** — `cwe` may be a string or an array; `advisory_ids` may be present or absent. Handle both.\n- **Sibling skill:** `vuln-scanner` produces the records this skill audits (its write actions — open PR, submit PVR, queue draft — are deliberately NOT here). `vuln-scanner`'s re-submit arm probes repos *waiting to open* PVR; Arm B picks up once a PVR is submitted. `inbox-triage` catches inbound maintainer replies via the GitHub notification layer (complementary, not a duplicate of Arm A's branch-name lifecycle audit).\n- Arm A, Arm B, and Arm C all read `memory\u002Fpending-disclosures\u002F` from different angles; coordinate via the shared `memory\u002Ftopics\u002Fvuln-followup.md` dashboard to avoid duplicate escalation.\n","> **${var}** — Scope selector for the lifecycle poll:\n> - empty → **full lifecycle poll**: PR\u002Fadvisory status + PVR triage + disclosure-queue aging (default).\n> - `prs` (also `pr` \u002F `tracker`) → **Arm A only** — PR\u002Fadvisory status audit + stars-secured dashboard.\n> - `pvr` → **Arm B only** — PVR triage-state poll on submitted advisories.\n> - `queue` (also `disclosures` \u002F `backlog`) → **Arm C only** — pending-disclosure queue aging + escalation.\n> - a bare `GHSA-xxxx-xxxx-xxxx` → **Arm B, single-advisory mode** — check just that one advisory's triage state on demand.\n\nToday is ${today}. This skill is the daily read\u002Fpoll arm of the vuln pipeline: `vuln-scanner` opens PRs, submits PVRs, and queues disclosure drafts, then moves on. This skill polls everything it produced and surfaces what the operator must look at — it does not open PRs or submit advisories itself (those are `vuln-scanner`'s write actions).\n\n## Voice\n\nIf `soul\u002FSOUL.md` and `soul\u002FSTYLE.md` are populated, read them and match the operator's voice in every notification. If empty or absent, use a clear, direct, neutral tone.\n\n## Capability mode\n\nThis skill runs `mode: write` deliberately. It is a read\u002Fpoll arm, but three of its capabilities cannot run under `read-only`:\n- **Arm B polls private, unpublished advisory triage state** via `gh api repos\u002F$REPO\u002Fsecurity-advisories\u002F$GHSA`. Draft\u002Ftriage advisories are visible only to the repo maintainers and the reporter, so the read is intrinsically authenticated; `read-only` strips `gh`, and a bare `$SECRET` on the command line is refused by the Bash permission layer (so a hand-rolled authenticated `curl` isn't an option either) — the read needs `gh api`, which handles auth internally.\n- **Arm B persists state transitions** — it rewrites `state:`\u002F`last_checked:`\u002F`resolved_at:` frontmatter in `memory\u002Fpending-disclosures\u002F*.md` in place and **moves** resolved files to `memory\u002Fpending-disclosures\u002Fresolved\u002F` (`Edit`\u002F`git mv` — both stripped in `read-only`).\n- **Arm A leans on authenticated `gh api`** for the PVR-state endpoint (`repos\u002F$REPO\u002Fprivate-vulnerability-reporting`) and repo\u002Fadvisory reads.\n\nNo sibling writes a repo file **outside** `memory\u002F`, so there was nothing to relocate — the only outside-`memory\u002F` writes are the ephemeral `.pending-notify-temp\u002F` notify-staging files. Keeping `write` preserves every absorbed capability.\n\n## Shared preamble (run for every invocation)\n\n1. Read `memory\u002FMEMORY.md` for context.\n2. Read the last ~3 days of `memory\u002Flogs\u002F` and drop anything already reported — don't re-surface the same signal twice.\n3. Read `soul\u002FSOUL.md` + `soul\u002FSTYLE.md` if populated (voice).\n4. **Parse `${var}` → scope** (deterministic; trim + lowercase, except a `GHSA-` value which is compared case-insensitively but preserved verbatim):\n   - empty → `scope = full` (run Arm A, then B, then C).\n   - matches `^GHSA-` (case-insensitive) → `scope = pvr`, `single_advisory = \u003Cthe GHSA value>` (Arm B filtered to one advisory).\n   - `prs` \u002F `pr` \u002F `tracker` → `scope = prs` (Arm A only).\n   - `pvr` → `scope = pvr` (Arm B only, all advisories).\n   - `queue` \u002F `disclosures` \u002F `backlog` → `scope = queue` (Arm C only).\n   - anything else → log `VULN_TRACKER_BAD_VAR: unrecognized scope '\u003Cvar>'`, send no notification, exit.\n5. `mkdir -p .pending-notify-temp` and start an empty combined-notification buffer at `.pending-notify-temp\u002Fvuln-tracker-${today}.md`. Each arm that has signal **appends its section** to this buffer; at the very end (step \"Notify\") the skill sends the buffer **once** if it is non-empty. This keeps notifications tight — a full poll with signal in two arms is one message, not two.\n\nThen run the arm(s) selected by `scope`, and finish with the shared **Notify** and **Log** steps.\n\n## Network Note\n\n- **Arm A & Arm B (GitHub reads):** all data via `gh api` \u002F `gh search` \u002F `gh pr view`. `gh` handles auth internally via `GH_TOKEN` (and Arm B's private-advisory reads need the elevated `GH_GLOBAL` PAT). No env-var-authenticated `curl` from bash — a bare `$SECRET` on the command line is refused by the Bash permission layer, so `gh api` (auth handled internally) is the reliable path; no postprocess scripts needed. Arm B keeps a documented `curl` fallback for the advisory endpoint — see Arm B step B2 — but `gh api` is preferred.\n- **Arm C (local only):** reads only local files (`memory\u002Fpending-disclosures\u002F`, `memory\u002Fissues\u002F`, `memory\u002Ftopics\u002Fpr-status.md`). No outbound network or auth required.\n- **Shell shape.** Each Bash call is one allowed command (pipes are fine). The run refuses a call that sets a variable (`BOT_AUTHOR=\"...\"`, `REPO=\"owner\u002Frepo\"`, `PVR_NOW=$(gh api ...)`) or joins commands with `;`, `&&` or `||`, and each refusal costs a model turn. So write the bot author, each repo, PR number and GHSA ID into each `gh` \u002F `.\u002Fsecretcurl` call literally (the `\u003CBOT_AUTHOR>`, `\u003CREPO>`, `\u003CNUM>`, `\u003CGHSA>` placeholders), make one call per PR, repo or advisory instead of a loop, and turn each `|| echo fallback` into a prose rule for when the call fails. Append each arm's section to the notify buffer with Edit (Write when the file is new), never with a shell `>>`.\n\n---\n\n## Arm A — PR & advisory lifecycle status  (scope `full` or `prs`)\n\nAudit the lifecycle status of every disclosure `vuln-scanner` has produced. Without a follow-up loop, three things rot silently:\n\n- **Merged-but-uncelebrated wins** — landed fixes never reach self-improve \u002F reflect (retro) without manual aggregation.\n- **Maintainer questions on open PRs** — a maintainer comments asking for clarification; if the bot doesn't see it, the PR ages out.\n- **Queued drafts past their disclosure window** — entries with `channel: \"skipped\"` (no-safe-channel) vanish into `vuln-scanned.json` with no recurring re-probe.\n\nThis arm cross-references `memory\u002Fvuln-scanned.json` against live GitHub state and surfaces anything the operator should look at.\n\n### A1. Load the canonical scan history\n\n**`memory\u002Fvuln-scanned.json` may be a flat top-level array of scan rows, not a `{scans: [...]}` object** - vuln-scanner writes the flat shape. Read it in a way that accepts both (the flat `[ {repo,scanned_at,findings,channel}, ... ]` shape and the legacy `{scans: [...]}` object), one `jq` call:\n\n```bash\njq -c 'if type==\"array\" then .[] else (.scans[]? \u002F\u002F empty) end' memory\u002Fvuln-scanned.json 2>\u002Fdev\u002Fnull\n```\n\nA bare `jq -c '.scans[]'` **errors** on a flat-array file (`Cannot index array with string \"scans\"`) and, with `2>\u002Fdev\u002Fnull` swallowing the error, silently yields zero rows - so the arm reads an empty scan history and reports only what other passes happen to surface. The `if type==\"array\"` guard is the fix; it keeps the legacy object shape working too.\n\nIf `memory\u002Fvuln-scanned.json` doesn't exist or the parsed row count is 0, log `VULN_TRACKER_SKIP: no scan history` for this arm and skip Arm A (no notification section - first runs of `vuln-scanner` haven't happened yet). In a `full` poll, continue to Arm B\u002FC.\n\nEach scan entry has at minimum: `repo`, `scanned_at`, `findings`, `channel`, `severity`. Public-PR entries also have `pr` (URL). Pending-disclosure entries have `draft_at` and `patch_branch`. Skipped entries have `reason`.\n\n**Retro-active coverage:** if the JSON was written after vuln-scanner started running, some PRs won't be in the JSON. Pull all bot-authored security PRs from GitHub directly (next step) to fill the gap.\n\n### A2. Pull all bot-authored security PRs from GitHub\n\n`vuln-scanner` opens security-fix PRs on a **`security\u002F`** branch. The **title** varies: dependency-bump PRs use **`fix(deps):`** (the A5a template is `fix(deps): bump \u003Cpkg> to patch \u003CCVE>`), and older or hand-written ones use **`fix(security):`** or a bare **`security:`**. **Match all three title prefixes, and treat the `security\u002F` branch prefix as the real invariant** - the branch name is mandated by A5a, the title wording is not.\n\n> **Do not filter on `fix(security):` alone.** vuln-scanner's dominant output is `fix(deps):`, so a `fix(security):`-only filter silently drops most in-flight security PRs - that is how `alibaba\u002Fopen-code-review#541` (conflicted, lint-failing, one unaddressed review) was missed by a poll while other PRs were tracked. Also exclude `docs(security):` - those are `security-sync` website PRs, not disclosures.\n\n`gh search prs --json` does **not** expose `headRefName` - that field is only available via GraphQL. So run **both** passes below every time and union them on `repository + number`; neither alone is complete.\n\nThe bot author is whoever the workflow uses (typically `github-actions[bot]` or a dedicated account configured in the workflow). Determine the author from `aeon.yml` or the workflow file; default to whatever account opened the most recent security-fix PR (any of the prefixes above) you can find.\n\n**Pass 1 (title prefixes, works everywhere)**, with the resolved bot author written in literally as `\u003CBOT_AUTHOR>`:\n\n```bash\ngh search prs --author \"\u003CBOT_AUTHOR>\" --json number,title,url,state,createdAt,closedAt,repository --limit 200 \\\n  | jq '[.[] | select(\n      (.title | startswith(\"fix(deps):\")) or\n      (.title | startswith(\"fix(security):\")) or\n      (.title | startswith(\"security:\"))\n    )]'\n```\n\n**Pass 2 (GraphQL, required - picks up every `security\u002F` branch whatever the title says)**, same author written in literally:\n\n```bash\ngh api graphql -f query='\n{\n  search(query: \"author:\u003CBOT_AUTHOR> is:pr sort:created-desc\", type: ISSUE, first: 100) {\n    nodes { ... on PullRequest {\n      number title url state createdAt closedAt mergedAt\n      repository { nameWithOwner }\n      headRefName\n    }}\n  }\n}' | jq '[.data.search.nodes[]\n    | select((.headRefName \u002F\u002F \"\") | startswith(\"security\u002F\"))]'\n```\n\nUnion the two result sets, dedup by `repository + number` (equivalently, by URL).\n\nCross-reference with `vuln-scanned.json`:\n- PR present in JSON → use JSON's `severity` \u002F `cwe` \u002F `note` for the row.\n- PR not in JSON → mark as `pre-history`; fill severity from the PR title if obvious.\n- JSON entry with `channel != \"public-pr\"` → no PR to fetch; goes in the \"queued\" \u002F \"skipped\" sections.\n\n### A3. Fetch live state for each open PR\n\nFor each open PR (state from step A2), one call per PR, with `\u003CREPO>` and `\u003CNUM>` written in literally:\n\n```bash\ngh pr view \"\u003CNUM>\" -R \"\u003CREPO>\" --json state,merged,closedAt,createdAt,reviews,comments,reviewDecision,author\n```\n\nPer-PR signals:\n- **Maintainer-needs-answer**: any comment whose `author.login != $BOT_AUTHOR` posted **after** the most recent comment by `$BOT_AUTHOR` (or after PR creation if the bot hasn't commented). Also `reviewDecision == \"CHANGES_REQUESTED\"` always counts.\n- **Stale-no-review**: `state == \"OPEN\"` AND no review AND no maintainer comment AND `createdAt` > 7d ago.\n- **Aging-with-engagement**: `state == \"OPEN\"` AND any maintainer activity AND open > 14d.\n\nIf a scan entry has `advisory_ids` (one or more GHSA IDs), check each one's published state, one call per GHSA ID with the origin repo and ID written in literally:\n\n```bash\ngh api \"repos\u002F\u003CORIGIN_REPO>\u002Fsecurity-advisories\u002F\u003CGHSA_ID>\" --jq '.state \u002F\u002F \"not found\"'\n```\n\n`state == \"published\"` → public advisory visible. 404 → osv-scanner referenced it but the upstream repo never published its own advisory.\n\n### A4. Fetch star counts for every secured repo\n\nFor every unique `repo` across the union from step A2 (JSON history + bot-authored security PRs), one call per repo with the repo written in literally:\n\n```bash\ngh api \"repos\u002F\u003CREPO>\" --jq '{stars: .stargazers_count, archived: .archived}' 2>\u002Fdev\u002Fnull\n```\n\n**Refetch every run.** Do NOT carry star counts forward from the previous `memory\u002Ftopics\u002Fvuln-followup.md` — per-repo counts drift between runs and the secured-stars headline is the operator's load-bearing metric. Cache only within a single run, keyed by `nameWithOwner`, so a repo with multiple PRs is fetched once.\n\nRepo-state handling:\n- **200 with stars**: use `.stargazers_count` (raw integer).\n- **200 with `archived: true`**: still use the star count, but suffix the repo cell with ` (archived)` so the operator knows the maintainer isn't responsive.\n- **404 \u002F 403**: repo was deleted, renamed, or made private. Record `null` and render as `repo-deleted`. Exclude from `total_stars_*` aggregates entirely so dead repos don't quietly zero out the totals.\n- **Other non-2xx**: record `null` and render `★?`. Flag in the run log for operator follow-up.\n\nThese per-repo counts power both the **Stars Secured** aggregate (step A6) and the `Stars` column on every per-repo table in step A6.\n\n### A5. Re-probe `channel: \"skipped\"` and `channel: \"pending-disclosure\"` repos\n\nFor each historical entry where the disclosure couldn't ship, re-check whether the situation changed:\n\n- **`channel: \"skipped\"`** with `reason` containing \"no PVR\":\n  ```bash\n  gh api \"repos\u002F\u003CREPO>\u002Fprivate-vulnerability-reporting\" --jq .enabled\n  ```\n  It prints PVR_NOW. If the call fails, PVR_NOW is `false`. If `PVR_NOW=true` and the original was `false`, surface as **newly-actionable**.\n- **`channel: \"skipped\"`** with `reason` containing \"no SECURITY.md\" - re-check `gh api repos\u002F\u003CREPO>\u002Fcontents\u002FSECURITY.md` and `.github\u002FSECURITY.md`. If now present, surface as **newly-actionable**.\n- **`channel: \"pending-disclosure\"`** — cross-reference with `memory\u002Fpending-disclosures\u002F` to see if the draft is still on disk. If the file is gone but the JSON entry says `pending-disclosure`, mark as `lost-draft` so it stops being escalated forever.\n\n### A6. Categorize every entry, then rewrite the dashboard\n\n| Status | Meaning |\n|---|---|\n| `merged` | PR merged. One-time celebration — drop from notifications after 30d. |\n| `open-clean` | PR open, no maintainer activity yet, \u003C 7d old. Wait. |\n| `needs-answer` | Maintainer commented or requested changes. **Operator action.** |\n| `stale-no-review` | Open > 7d, zero maintainer activity. Consider polite ping or close. |\n| `aging-engaged` | Open > 14d with engagement. Operator should triage. |\n| `closed-no-merge` | PR closed without merging. Capture the reason for review. |\n| `queued` | `pending-disclosure` draft on disk, not yet shipped. |\n| `skipped-rechecked` | Channel was \"skipped\" originally; re-probe still shows no channel. |\n| `newly-actionable` | Skipped originally; PVR or SECURITY.md now present. **Operator action.** |\n| `lost-draft` | JSON says pending-disclosure but draft file is gone. Display once, then suppress. |\n| `pre-history` | PR found via search but predates `vuln-scanned.json`. Fill what we can. |\n\nThen rewrite `memory\u002Ftopics\u002Fvuln-followup.md` (rewrite — don't append; this file is a living dashboard, not a log).\n\nThe **Stars Secured** block goes at the top so the operator sees aggregate impact before drilling into rows. `total_stars_secured` = sum of stargazers across every unique repo where vuln-scanner has landed at least one merged PR. `total_stars_in_flight` = sum across repos with an open PR. `total_stars_tracked` = sum across the full union. Track all three because celebration uses `secured`, prioritization uses `in_flight`, and historical review uses `tracked`.\n\nRound star counts to abbreviated form for the headline (12.4k, 1.8k, 940). Keep raw integers in the per-repo tables so sort\u002Fdiff stays exact.\n\n```markdown\n# Vuln Tracker Status\n\n*Last updated: ${today}*\n\n## Stars Secured\n\n- **Merged-PR repos (secured):** ★ \u003Ctotal_stars_secured> across \u003Csecured_repo_count> repos\n- **Open-PR repos (in flight):** ★ \u003Ctotal_stars_in_flight> across \u003Cin_flight_repo_count> repos\n- **All tracked repos:** ★ \u003Ctotal_stars_tracked> across \u003Ctotal_repo_count> repos\n\n### Secured leaderboard - every merged PR ranked by repo stars\n| Rank | Repo | Stars | PR | Merged | Severity | Title |\n|------|------|-------|----|--------|----------|-------|\n\n### Per-repo breakdown - secured (sorted by stars desc)\n| Repo | Stars | Merged PRs | First merge | Latest merge | Severities landed |\n|------|-------|------------|-------------|--------------|-------------------|\n\n### Per-repo breakdown - in flight (sorted by stars desc)\n| Repo | Stars | Open PRs | Oldest open | Severities open |\n|------|-------|----------|-------------|-----------------|\n\n### Per-repo breakdown - queued \u002F skipped \u002F closed (sorted by stars desc)\n| Repo | Stars | Status | Severity | Note |\n|------|-------|--------|----------|------|\n\n## Operator-action queue\n\n### Needs answer (\u003Ccount>)\n| Repo | Stars | PR | Title | Last activity | Latest commenter |\n|------|-------|----|----|--------------|------------------|\n\n### Newly actionable - channel opened up since the original scan (\u003Ccount>)\n| Repo | Stars | Original date | Original blocker | Now |\n|------|-------|---------------|------------------|-----|\n\n### Stale or aging\n| Repo | Stars | PR | Age | Status | Suggested action |\n|------|-------|----|----|-----|------------------|\n\n## Recently merged (last 30d, \u003Ccount>)\n| Date merged | Repo | Stars | PR | Severity | Title |\n|-------------|------|-------|----|----------|-------|\n\n## Open \u002F clean (no operator action - wait, \u003C 7d) (\u003Ccount>)\n| Repo | Stars | PR | Severity | Opened | Age |\n|------|-------|----|----------|--------|-----|\n\n## Closed without merge (last 30d, \u003Ccount>)\n| Date | Repo | Stars | PR | Severity | Title | Likely reason |\n|------|------|-------|----|----------|-------|---------------|\n\n## Queued (no PR yet) (\u003Ccount>)\n| Severity | Repo | Stars | Original channel | Original blocker | Days queued |\n|----------|------|-------|------------------|------------------|-------------|\n\n## Lost-draft ghosts (suppressed from notifications)\n| Date | Repo | Stars | Severity |\n```\n\n### A7. Decide whether Arm A has signal\n\n**No signal (contribute nothing to the notification)** if all are true:\n- Zero `needs-answer`\n- Zero `newly-actionable`\n- Zero items moved to `merged` since the last run\n- Zero items moved to `closed-no-merge` since the last run\n- Zero items aged into `stale-no-review` or `aging-engaged` since the last run\n\nTo detect \"since last run,\" diff today's categorization against the previous `memory\u002Ftopics\u002Fvuln-followup.md`. If the file doesn't exist (first run), treat all entries as new and surface the full backlog.\n\n### A8. Emit Arm A's notification section (when it has signal)\n\nAppend this section to the combined buffer (`.pending-notify-temp\u002Fvuln-tracker-${today}.md`). Keep the whole message under 4000 chars; if tight, drop the merged\u002Fopened\u002Fstale section bodies and keep counts only.\n\n```\n*PR & advisory status*\n\n★ secured: \u003Ctotal_stars_secured> across \u003Csecured_repo_count> repos  (in flight: \u003Ctotal_stars_in_flight> \u002F \u003Cin_flight_repo_count>)\n\nneeds answer: \u003CN>\n\u003Crepo> (★\u003Cstars>) #\u003Cnum> - \u003Clatest_commenter>: \"\u003Ccomment_excerpt_first_120_chars>\"\n\nnewly actionable: \u003CN>\n\u003Crepo> (★\u003Cstars>, queued \u003Cdays>d) - \u003Cwhat_changed>\n\nmerged this week: \u003CN>  \u003Clist_with_stars>\nopened, waiting: \u003CN>\nstale: \u003CN>\nqueued: \u003CN> (\u003Ccritical>C \u002F \u003Chigh>H \u002F \u003Cother>M+L)\n\nleaderboard top-3 (PRs by ★): #1 \u003Crepo1> ★\u003Cs1> (PR #\u003Cpr1>) - #2 \u003Crepo2> ★\u003Cs2> (PR #\u003Cpr2>) - #3 \u003Crepo3> ★\u003Cs3> (PR #\u003Cpr3>)\n\ndashboard: memory\u002Ftopics\u002Fvuln-followup.md (full leaderboard inside)\n```\n\nRecord Arm A status `VULN_TRACKER_OK` for the log.\n\n---\n\n## Arm B — PVR triage state  (scope `full`, `pvr`, or a bare `GHSA-…`)\n\n`vuln-scanner`'s re-submit arm (`var=resubmit`, the former `pvr-watchlist`) monitors repos *waiting to open* PVR. This arm monitors PVRs that have **already been submitted** and tracks their lifecycle: `triage` → `draft` (accepted) → `published` (public) or `withdrawn` (rejected). Without this, submitted advisories sit unmonitored until manually recalled from memory.\n\nSource of truth: `memory\u002Fpending-disclosures\u002F*.md` files with `channel: pvr` frontmatter. Each such file must have `ghsa`, `repo`, `state`, `submitted_at` fields (full schema below).\n\n**Configuration — optional tracking issue.** The arm can reference an optional **tracking issue** in the operator's own repo — useful for cross-linking advisory state with an internal issue board. Resolve from (priority order):\n1. `aeon.yml` top-level key `pvr_triage.tracking_issue:` (e.g. `pvr_triage: { tracking_issue: \"owner\u002Frepo#123\" }`)\n2. environment variable `AEON_PVR_TRACKING_ISSUE`\n3. unset → skip cross-linking entirely.\n\nIf a tracking issue is configured, mention its URL in the notification and the per-advisory write-up so the operator can navigate to the canonical tracker.\n\n### B1. Discover in-flight PVRs\n\nScan `memory\u002Fpending-disclosures\u002F` for all `.md` files. Parse the YAML frontmatter. Keep only those with `channel: pvr`.\n\nIf `single_advisory` is set (a bare `GHSA-…` was passed), filter to just the matching `ghsa` value (one-off mode).\n\nIf no PVR files found:\n```\nPVRT_SKIP: no submitted PVRs on disk\n```\nRecord that Arm B status and skip Arm B (no notification section). In a `full` poll, continue to Arm C.\n\n### B2. Probe each advisory's triage state\n\nFor each entry, determine `repo` and `ghsa` from frontmatter. Then one call per entry, with them written in literally (for example `owner\u002Frepo` and `GHSA-xxxx-xxxx-xxxx`):\n\n```bash\ngh api \"repos\u002F\u003CREPO>\u002Fsecurity-advisories\u002F\u003CGHSA>\" \\\n  --jq '{state: .state, cve_id: .cve_id, published_at: .published_at}' 2>&1\n```\n\nExpected outcomes:\n\n| Response | Meaning |\n|----------|---------|\n| `{state: \"triage\", ...}` | Maintainer hasn't reviewed yet |\n| `{state: \"draft\", ...}` | Accepted — maintainer is working on it |\n| `{state: \"published\", ...}` | Published — fully resolved |\n| `{state: \"withdrawn\", ...}` | Rejected or withdrawn by reporter |\n| HTTP 403 | Private advisory, we don't have read access — state unknown, treat as still `triage` |\n| HTTP 404 | Advisory deleted \u002F repo private \u002F GHSA invalid — flag as `not-found` |\n\n**Fallback:** `gh api` uses `GH_TOKEN` internally (workflow wires `GH_GLOBAL` for the elevated advisory read). If `gh` is somehow unavailable, fall back to `.\u002Fsecretcurl` with the `{GH_TOKEN}` placeholder (same token; `GH_GLOBAL` itself is not a substitutable name):\n```bash\n.\u002Fsecretcurl -s -H \"Authorization: Bearer {GH_TOKEN}\" \\\n  \"https:\u002F\u002Fapi.github.com\u002Frepos\u002F\u003CREPO>\u002Fsecurity-advisories\u002F\u003CGHSA>\" \\\n  | grep -o '\"state\":\"[a-z]*\"'\n```\n(Never write a raw `curl` with `$GH_GLOBAL` \u002F `$GH_TOKEN`: a bare `$SECRET` on the command line is refused by the Bash permission layer. `gh api` stays the primary path.)\n\n### B3. Detect state changes\n\nCompare the probed `state` to the `state` in the frontmatter.\n\n- **No change:** note it, continue.\n- **Changed:** this is the primary event. Log old → new state.\n\nAlso flag:\n- **Aged triage:** `state=triage` AND (`today` − `submitted_at`) > 30 days → escalate. Most maintainers respond within 30 days; silence past that is actionable.\n- **Accepted (draft):** surface the patch branch from the `patch_branch` frontmatter field — maintainer may want a PR instead of a private advisory.\n- **Published:** advisory is live. The finding is closed. Update state and mark for removal.\n- **Withdrawn:** rejected. Note the reason if visible. Mark for cleanup.\n\n### B4. Update frontmatter in-place\n\nFor each file with a state change, rewrite just the `state` field in the YAML frontmatter. Also update a `last_checked` field (add it if absent).\n\nDo NOT modify the body of the advisory file — only update frontmatter.\n\nExample frontmatter update:\n```yaml\nstate: draft          # was: triage\nlast_checked: 2026-05-21\n```\n\nFor `published` or `withdrawn` entries, add:\n```yaml\nresolved_at: 2026-05-21\n```\n\n### B5. Decide whether Arm B has signal\n\n- **All entries still `triage`, no changes, none aged:** no notification section. Record status and continue.\n- **Any state change, aged entry, or action item:** emit the section below.\n\n### B6. Emit Arm B's notification section (when it has signal)\n\nAppend to the combined buffer:\n\n```\n*PVR triage*\n\npvr triage: {total} advisories in flight. {changed_count} changed.\n\nCHANGED:\n- {repo} {ghsa} - {old_state} → {new_state}\n  {action_item}\n\nAGED (>30d no response):\n- {repo} {ghsa} - {days}d in triage. {severity}. escalate or close.\n  patch: {patch_branch}\n\nSTILL TRIAGE:\n{n} advisories waiting. oldest: {repo} ({days}d).\n\n{if tracking_issue configured}\ntracker: {tracking_issue_url}\n{end}\n```\n\nAction items by transition:\n- `triage → draft` → \"maintainer accepted — offer to PR the patch branch: {patch_branch}\"\n- `triage → published` → \"published as {cve_id}. remove from tracking.\"\n- `triage → withdrawn` → \"rejected. remove from tracking and note in vuln-scanned.json.\"\n- aged triage (>30d) → \"30d+ no response. consider pinging maintainer or withdrawing.\"\n\n### B7. Clean up resolved entries\n\nFor entries where `state=published` or `state=withdrawn` AND `resolved_at` is set: move the file from `memory\u002Fpending-disclosures\u002F` to `memory\u002Fpending-disclosures\u002Fresolved\u002F` (create the directory if needed; use `git mv`).\n\nDo NOT delete — keep as a historical record.\n\nRecord Arm B status `PVRT_OK` for the log.\n\n---\n\n## Arm C — Disclosure-queue aging  (scope `full` or `queue`)\n\nMonitor the pending vulnerability disclosure backlog. `vuln-scanner` queues draft advisories to `memory\u002Fpending-disclosures\u002F` when PVR auto-submission fails or the disclosure path is email-only. Without daily visibility, CRITICAL\u002FHIGH advisories silently age past responsible-disclosure windows. This arm surfaces the queue state and escalates when findings sit too long. (Arm B tracks the lifecycle of the `channel: pvr` subset; Arm C ages the **whole** queue by severity and disclosure status.)\n\n### C1. Scan the backlog\n\n```bash\nls memory\u002Fpending-disclosures\u002F 2>\u002Fdev\u002Fnull\n```\n\nIf the directory doesn't exist or is empty: log `DISCLOSURE_TRACKER_SKIP: no pending advisories` for this arm and skip Arm C (no notification section).\n\n### C2. Parse each advisory file\n\nFor each `.md` file in `memory\u002Fpending-disclosures\u002F` (ignore the `resolved\u002F` subdir):\n\n**From the filename** (pattern: `{repo-slug}-{YYYY-MM-DD}.md` or `{repo-slug}-{YYYY-MM-DD}-{ampm}.md`):\n- Extract target repo slug (everything before the last date segment)\n- Extract filed date\n\n**From the YAML frontmatter** (if present) parse:\n- `repo:` — overrides the filename slug when present (canonical target)\n- `severity:` — CRITICAL \u002F HIGH \u002F MEDIUM \u002F LOW\n- `status:` — see step C2.5 for the controlled vocabulary\n\n**From the file content** (fallback for files without frontmatter), look for these near the top of the file:\n- `Severity:` or `**Severity:**` — one of CRITICAL \u002F HIGH \u002F MEDIUM \u002F LOW\n- `CVE\u002FCWE:` or similar identifier\n- Short title (first non-blank heading line)\n\nIf severity is not parseable, treat as MEDIUM. Compute age: `today - filed date` in days.\n\n### C2.5. Classify each advisory's disclosure state\n\nBefore counting any draft as a past-threshold escalation, decide whether the draft is genuinely pending or already covered. A draft can be in one of these states:\n\n- `escalate` — pending, no canonical PR found, past the severity-tier threshold\n- `pending` — pending, no canonical PR found, within the threshold window\n- `operator-todo` — needs operator-only action (email send, PVR enable nudge); not an agent failure\n- `covered-by-pr` — a canonical disclosure PR has already been filed against the target repo and is `OPEN` or recently merged\n- `superseded-upstream` — the bypass \u002F vuln is fixed in upstream already, draft is dead-weight\n- `submitted` — already submitted via PVR \u002F GHSA; awaiting maintainer response\n\nResolution rules:\n\n1. **Check frontmatter `status:` first** — map the literal value to a state:\n   - `superseded-upstream` → `superseded-upstream`\n   - `submitted`, `submitted-via-pvr`, `disclosed-via-pr-{N}`, `email-sent` → `submitted` or `covered-by-pr`\n   - `pending-operator-send` **with `auto_send: true`** → `pending` (`vuln-scanner`'s disclose arm, `var=disclose`, sends it autonomously - not a human task)\n   - `pending-operator-send` (without `auto_send: true`), `queued for operator manual send`, `email-failed`, any string mentioning \"operator\" → `operator-todo`\n   - `pending`, blank, or missing → fall through to rule 2\n\n2. **Cross-reference `memory\u002Ftopics\u002Fpr-status.md`** (if present) — grep for the `{repo}` slug (frontmatter `repo:` or filename) in the Open section and Recent Merges section. If a row exists with a `fix(security)` or `chore(security)` title against that repo, opened on or after the draft's `detected_at` \u002F `reconstructed_at` \u002F filed-date, classify as `covered-by-pr` and capture the PR number \u002F title for the summary. If `memory\u002Ftopics\u002Fpr-status.md` doesn't exist, skip this lookup and fall to rule 3.\n\n3. **Fall through** — if no status hint and no canonical PR found, classify as `pending`. Then check age vs the severity-tier threshold (CRITICAL 3d \u002F HIGH 7d \u002F MED-LOW 14d) — if past, promote to `escalate`.\n\nThis is the load-bearing step. Without cross-referencing already-merged fix PRs the tracker generates false-positive escalations for drafts that have already been resolved.\n\n### C3. Build the summary\n\nGroup advisories by **state first**, then by severity within each state. The three buckets are:\n\n- **Escalate** — `escalate` state only (truly stuck, past threshold, no canonical PR)\n- **Operator-todo** — `operator-todo` state (email-only sends, PVR-enable nudges, anything awaiting human action)\n- **Cleanup candidates** — `covered-by-pr`, `submitted`, `superseded-upstream` (draft files that can be removed from `memory\u002Fpending-disclosures\u002F`)\n\nSeverity tiers and thresholds (only apply to `escalate` and `pending` states):\n- **CRITICAL** (age threshold: 3 days — escalate immediately)\n- **HIGH** (age threshold: 7 days — escalate at 7d)\n- **MEDIUM \u002F LOW** (threshold: 14 days)\n\nFor each advisory, produce one line:\n```\n- {repo-slug} | {severity} | {age}d | {short title}{state-suffix}\n```\nwhere `{state-suffix}` is empty for `escalate` \u002F `pending`, `[operator-todo: {reason}]` for operator-todo, and `[covered: PR #{N}]` \u002F `[superseded-upstream]` \u002F `[submitted]` for cleanup candidates.\n\nCount totals. Identify advisories in the `escalate` state — those are the only ones that drive the urgent notification path.\n\n### C4. Check for upstream PVR \u002F token issues\n\nLook in `memory\u002Fissues\u002FINDEX.md` for any open issues tagged with `pvr`, `repository_advisories`, or `missing-secret` that explain why advisories are stuck. If such an issue exists, note:\n- Number of consecutive PVR failures (if logged)\n- Fix estimate from the issue notes\n- How many of the backlogged advisories are blocked by it\n\nIf no such issue exists, treat the queue as routine and skip the \"blocked by\" line in the notification.\n\n### C5. Decide whether Arm C has signal\n\nCompute counts from step C3:\n- `escalate_count` — drafts in the `escalate` state\n- `pending_count` — drafts in `pending` (in-window) state\n- `operator_todo_count` — drafts awaiting operator action\n- `cleanup_count` — drafts in `covered-by-pr` \u002F `submitted` \u002F `superseded-upstream`\n\nDecision:\n- **Queue empty**: log `DISCLOSURE_TRACKER_SKIP: queue empty`, no section.\n- **`escalate_count` > 0**: emit the **urgent** section.\n- **`escalate_count` == 0 but `cleanup_count` > 0**: emit the **daily digest** section (includes the cleanup-candidate list so the operator can prune `memory\u002Fpending-disclosures\u002F`).\n- **All `pending` \u002F `operator-todo`, nothing past threshold, no cleanup candidates**: emit the **daily digest** section.\n\nCoverage from `covered-by-pr` \u002F `submitted` \u002F `superseded-upstream` is **never** counted as an escalation — informational only. Operator-todo is **never** an escalation either; it's surfaced separately so the operator knows their inbox.\n\n### C6. Emit Arm C's notification section\n\nAppend to the combined buffer.\n\n**Urgent format** (`escalate_count` > 0):\n\n```\n*Disclosure queue*\n\ndisclosure queue: {escalate_count} past threshold (of {total} drafts).\n\nESCALATE:\n- {repo} - {severity}, {age}d old (threshold: {N}d)\n[... others in `escalate` state ...]\n\noperator-todo ({operator_todo_count}):\n- {repo} - {severity}, {age}d - {operator-reason}\n\ncleanup candidates ({cleanup_count}):\n- {repo} - [covered: PR #{N} \u002F superseded-upstream \u002F submitted] - safe to delete from memory\u002Fpending-disclosures\u002F\n\n{IF blocking issue tracked in memory\u002Fissues\u002FINDEX.md}\nblocked by {ISS-ID} - {short reason}\nfix: {fix estimate} unblocks {N} of {escalate_count + pending_count}\n{end}\n```\n\n**Daily digest format** (no escalation):\n\n```\n*Disclosure queue*\n\ndisclosure queue: {total} drafts. {critical_count} CRITICAL, {high_count} HIGH, {other_count} MED\u002FLOW.\n{pending_count} in-window, {operator_todo_count} operator-todo, {cleanup_count} cleanup candidates.\noldest in-window: {repo} ({age}d).\n{cleanup section if cleanup_count > 0}\n{IF blocking issue tracked in memory\u002Fissues\u002FINDEX.md}\nblocked by {ISS-ID} - {short reason}.\n{end}\n```\n\nRecord Arm C status `DISCLOSURE_TRACKER_OK` for the log.\n\n---\n\n## Notify (shared, runs once at the end)\n\nIf the combined buffer `.pending-notify-temp\u002Fvuln-tracker-${today}.md` has at least one arm section, prepend the header line `*Vuln Tracker — ${today}*` and send once:\n\n```bash\n.\u002Fnotify -f .pending-notify-temp\u002Fvuln-tracker-${today}.md\n```\n\nIf the buffer is empty (every arm that ran was silent), **send nothing** — a clean poll is silent. Keep the whole message under 4000 chars; if over, trim Arm A's body first (keep counts), then Arm C's cleanup list, keeping every `needs-answer` \u002F `newly-actionable` \u002F `ESCALATE` line.\n\n## Log (shared, runs once at the end)\n\nAppend to `memory\u002Flogs\u002F${today}.md` under ONE heading (the health loop parses `### \u003Cskill-name>`), with a discriminator line naming the scope\u002Farms that ran:\n\n```\n### vuln-tracker\n- scope: \u003Cfull | prs | pvr | queue | single-advisory:GHSA-…>\n- Arm A (PR\u002Fadvisory): \u003Cstatus> - \u003Ctotal_in_json> JSON + \u003Cpre_history_prs> pre-history = \u003Ctotal> tracked; states: \u003Cmerged> merged \u002F \u003Copen_clean> open-clean \u002F \u003Cneeds_answer> needs-answer \u002F \u003Cstale> stale \u002F \u003Caging> aging \u002F \u003Cclosed> closed \u002F \u003Cqueued> queued \u002F \u003Cskipped> skipped-rechecked \u002F \u003Cnewly_actionable> newly-actionable\n- Stars: ★\u003Ctotal_stars_secured> secured (\u003Csecured_repo_count> repos) \u002F ★\u003Ctotal_stars_in_flight> in flight (\u003Cin_flight_repo_count> repos) \u002F ★\u003Ctotal_stars_tracked> tracked (\u003Ctotal_repo_count> repos)\n- Arm B (PVR triage): \u003Cstatus> - checked \u003Ctotal>; changed \u003Cchanged_count> (\u003Clist>); aged>30d \u003Caged_count>; still-triage \u003Cwaiting_count>; tracking issue \u003Curl|none>\n- Arm C (disclosure queue): \u003Cstatus> - \u003Ctotal> drafts (\u003Ccritical_count> CRITICAL \u002F \u003Chigh_count> HIGH \u002F \u003Cother_count> MED\u002FLOW); \u003Cescalate_count> escalate \u002F \u003Cpending_count> in-window \u002F \u003Coperator_todo_count> operator-todo \u002F \u003Ccleanup_count> cleanup; oldest in-window \u003Crepo> (\u003Cage>d); blocking issue \u003CISS-ID|none>\n- Operator queue this run: \u003Cneeds_answer + newly_actionable + aged PVRs + escalate_count>\n- Notification: \u003Csent | skipped (no movement)>\n- VULN_TRACKER_OK\n```\n\nOnly include the Arm B \u002F Arm C lines for arms that actually ran under the current scope. Use each arm's own status token (`VULN_TRACKER_OK` \u002F `VULN_TRACKER_SKIP`, `PVRT_OK` \u002F `PVRT_SKIP`, `DISCLOSURE_TRACKER_OK` \u002F `DISCLOSURE_TRACKER_SKIP`) inline, and close with `VULN_TRACKER_OK` for the hub run (or `VULN_TRACKER_BAD_VAR` if the var was unrecognized).\n\n## Required Env Vars\n\n- `GH_TOKEN` \u002F `GITHUB_TOKEN` — GitHub read token for Arm A's PR\u002Fcomment\u002Frepo\u002Fadvisory reads and the PVR-state endpoint. `repo` scope is enough. Present by default in GitHub Actions.\n- `GH_GLOBAL` — GitHub PAT with `public_repo` + `repository_advisories:write` scope, used by Arm B to read the triage state of **private\u002Funpublished** advisories you submitted. Same token `vuln-scanner` uses (all arms). Without it, Arm B advisory reads may 403 (treated as still `triage`).\n- `AEON_PVR_TRACKING_ISSUE` (optional) — internal tracking-issue reference for Arm B cross-linking (else resolved from `aeon.yml` `pvr_triage.tracking_issue:`, else skipped).\n- Arm C requires no env vars — all its data comes from local files written by `vuln-scanner`.\n\n## Pending-disclosure file schema (Arm B & Arm C)\n\n`memory\u002Fpending-disclosures\u002F*.md` files. Newer drafts use YAML frontmatter:\n\n```yaml\n---\nrepo: owner\u002Frepo\nghsa: GHSA-xxxx-xxxx-xxxx\nghsa_url: https:\u002F\u002Fgithub.com\u002Fowner\u002Frepo\u002Fsecurity\u002Fadvisories\u002FGHSA-xxxx-xxxx-xxxx\nchannel: pvr                       # Arm B tracks only channel: pvr entries\nstate: triage                      # triage | draft | published | withdrawn\nstatus: pending-operator-send      # optional; Arm C vocabulary below\nsubmitted_at: 2026-05-12T19:54:42Z\nlast_checked: 2026-05-15           # added\u002Fupdated by Arm B\nseverity: HIGH\ncwe: [CWE-xxx]                      # may be a string or an array — handle both\npatch_branch: https:\u002F\u002Fgithub.com\u002F\u003Cfork-owner>\u002Frepo\u002Ftree\u002Fsecurity\u002Fbranch-name\npatch_commit: abc1234\nsubmit_url: https:\u002F\u002Fgithub.com\u002Fowner\u002Frepo\u002Fsecurity\u002Fadvisories\u002Fnew\n---\n\n# {Repo}: {Title}\n...\n```\n\n- **Arm B required fields:** `repo`, `ghsa`, `channel: pvr`, `state`, `submitted_at`. Optional: `patch_branch`, `patch_commit`, `cwe`, `ghsa_url`.\n- **Older drafts** use inline `**Severity:**` lines and no frontmatter — parse defensively: grep for `severity:` and `Severity:` case-insensitively; if unparseable, default to MEDIUM.\n\n### `status:` controlled vocabulary (Arm C, step C2.5)\n\nSet by `vuln-scanner` \u002F operator \u002F cleanup chores:\n\n- (blank or missing) — pending; falls through to the PR-status cross-ref\n- `pending` — same as blank\n- `pending-operator-send` **with `auto_send: true`** - queued for autonomous send by `vuln-scanner`'s disclose arm (`var=disclose`); classify as `pending` (NOT operator-todo). It flips to `email-sent` on the next disclose run, which only happens when `vuln-scanner` is dispatched with `var=disclose` (the default schedule runs the scan arm). If it lingers armed-but-unsent across several days, either no disclose run was dispatched or `RESEND_API_KEY` isn't configured - surface it once so the operator schedules or runs the disclose arm and wires up Resend.\n- `pending-operator-send` **with `auto_send: false`\u002Fabsent** \u002F `queued for operator manual send` — operator-todo (human must send; e.g. non-email contact or AI-report ban)\n- `email-sent` - sent by `vuln-scanner`'s disclose arm via Resend; awaiting maintainer reply. Treat like `submitted` (informational cleanup candidate, **never** an escalation).\n- `email-failed` - the disclose arm gave up after repeated send failures (bad address \u002F Resend error). **operator-todo** — the contact needs fixing or a manual send.\n- `submitted` \u002F `submitted-via-pvr` — submitted, awaiting maintainer\n- `disclosed-via-pr-\u003CN>` — covered-by-pr, draft can be archived\n- `superseded-upstream` — bypass is already fixed in upstream; draft is dead\n- Any string containing `operator` — operator-todo\n\nWhen a canonical PR lands but the draft's `status:` was never set, Arm C falls through to step C2.5 rule 2 (cross-ref `pr-status.md`) and classifies as `covered-by-pr` automatically. The `status:` shortcut just bypasses the grep.\n\n**When to delete a draft:** cleanup candidates in the notification can be removed by the operator with `rm memory\u002Fpending-disclosures\u002F\u003Cfile>.md`. Safe once the canonical PR is open — the patch branch on the fork remains the authoritative artifact.\n\n## Notes & related\n\n- **`vuln-scanned.json` schema is loose** — `cwe` may be a string or an array; `advisory_ids` may be present or absent. Handle both.\n- **Sibling skill:** `vuln-scanner` produces the records this skill audits (its write actions — open PR, submit PVR, queue draft — are deliberately NOT here). `vuln-scanner`'s re-submit arm probes repos *waiting to open* PVR; Arm B picks up once a PVR is submitted. `inbox-triage` catches inbound maintainer replies via the GitHub notification layer (complementary, not a duplicate of Arm A's branch-name lifecycle audit).\n- Arm A, Arm B, and Arm C all read `memory\u002Fpending-disclosures\u002F` from different angles; coordinate via the shared `memory\u002Ftopics\u002Fvuln-followup.md` dashboard to avoid duplicate escalation.\n",{"name":5,"metadata":44,"description":7},{"var":45,"mode":35,"tags":46,"title":48,"category":28,"requires":49,"depends_on":52},"",[47,32,24],"meta","Vuln Tracker",[50,51],"GH_TOKEN?","GH_GLOBAL?",[53],"vuln-scanner","main","504a21d53c0c9b485f6e90c43db83000825e1d3c","skills\u002Fvuln-tracker",308,"https:\u002F\u002Fgithub.com\u002Faeonfun\u002Faeon","https:\u002F\u002Fwww.aeon.fun\u002F","https:\u002F\u002Fgithub.com\u002Faeonfun","The most autonomous agent framework.","2026-11-15 03:40:18",{"source":8,"skillName":5,"commands":64,"prompt":74,"zipAvailable":75},[65,68,71],{"tool":66,"command":67},"npx","npx skills add aeonfun\u002Faeon --skill vuln-tracker",{"tool":69,"command":70},"bunx","bunx skills add aeonfun\u002Faeon --skill vuln-tracker",{"tool":72,"command":73},"pnpm","pnpm dlx skills add aeonfun\u002Faeon --skill vuln-tracker","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Faeonfun\u002Faeon\u002Fblob\u002Fmain\u002Fskills\u002Fvuln-tracker\u002FSKILL.md",true,[],[78,85,91,97,103,109],{"skillNo":79,"name":80,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":81,"categories":82,"keywords":83,"publishedAt":84},"01M7DSM3QWBJ44FKRMDDRAGNET","Warn Mixed Script","https:\u002F\u002Fgithub.com\u002Faeonfun\u002Faeon\u002Fblob\u002Fmain\u002Fscripts\u002Ftests\u002Fskill-scan-fixtures\u002Fwarn-mixed-script\u002FSKILL.md",[],[],"2026-11-15 14:19:17",{"skillNo":86,"name":87,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":88,"categories":89,"keywords":90,"publishedAt":84},"01M7DSM0496DRYYW5A4NAAHZDX","Warn Hidden CSS","https:\u002F\u002Fgithub.com\u002Faeonfun\u002Faeon\u002Fblob\u002Fmain\u002Fscripts\u002Ftests\u002Fskill-scan-fixtures\u002Fwarn-hidden-css\u002FSKILL.md",[],[],{"skillNo":92,"name":93,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":94,"categories":95,"keywords":96,"publishedAt":84},"01M7DSKWCGDVYZSDJM31YHS2JB","Malicious Injection Uppercase","https:\u002F\u002Fgithub.com\u002Faeonfun\u002Faeon\u002Fblob\u002Fmain\u002Fscripts\u002Ftests\u002Fskill-scan-fixtures\u002Fmalicious-injection-uppercase\u002FSKILL.md",[],[],{"skillNo":98,"name":99,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":100,"categories":101,"keywords":102,"publishedAt":84},"01M7DSKRV2RRA1V0HK13KYWJXD","Malicious Injection Homoglyph","https:\u002F\u002Fgithub.com\u002Faeonfun\u002Faeon\u002Fblob\u002Fmain\u002Fscripts\u002Ftests\u002Fskill-scan-fixtures\u002Fmalicious-injection-homoglyph\u002FSKILL.md",[],[],{"skillNo":104,"name":105,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":106,"categories":107,"keywords":108,"publishedAt":84},"01M7DSKMXJKY0167YHVCKF7B49","Malicious Injection Disregard Prior","https:\u002F\u002Fgithub.com\u002Faeonfun\u002Faeon\u002Fblob\u002Fmain\u002Fscripts\u002Ftests\u002Fskill-scan-fixtures\u002Fmalicious-injection-disregard-prior\u002FSKILL.md",[],[],{"skillNo":110,"name":111,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":112,"categories":113,"keywords":114,"publishedAt":84},"01M7DSKGFQ5V2VNSA4K97CVFMB","Benign Greek Math","https:\u002F\u002Fgithub.com\u002Faeonfun\u002Faeon\u002Fblob\u002Fmain\u002Fscripts\u002Ftests\u002Fskill-scan-fixtures\u002Fbenign-greek-math\u002FSKILL.md",[],[]]