[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6CZ7TXV2TMG5CQ0NJ2BXQQT":3,"skill-same:01M6CZ7TXV2TMG5CQ0NJ2BXQQT":70},{"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":38,"linkOnly":39,"markdownContent":40,"markdownBody":41,"frontMatter":42,"refName":47,"commitSha":48,"skillDirectory":49,"forksCount":50,"repositoryHtmlUrl":51,"repositoryHomepageUrl":52,"ownerHtmlUrl":53,"ownerBlogUrl":52,"ownerDescription":54,"updatedAt":55,"install":56,"resources":69},"01M6CZ7TXV2TMG5CQ0NJ2BXQQT","fork-fleet",null,"Fork divergence monitor - tracks where the fleet's active forks diverge in CODE (unique commits, new\u002Fmodified skills) and CONFIG (enable\u002Fvar\u002Fmodel\u002Fschedule vs upstream), gated on real change.","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\u002Ffork-fleet\u002FSKILL.md",[],[19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37],"fork","fleet","fork fleet","schedule","new","tracks","active","upstream","gated","diverge","change","model","commits","Meta","dev","code","core","monitor","config","2026-11-10 10:12:30",false,"---\nname: fork-fleet\ndescription: Fork divergence monitor - tracks where the fleet's active forks diverge in CODE (unique commits, new\u002Fmodified skills) and CONFIG (enable\u002Fvar\u002Fmodel\u002Fschedule vs upstream), gated on real change.\nmetadata:\n  category: core\n  var: \"\"\n  tags:\n    - dev\n    - meta\n---\n> **${var}** — Divergence scope selector; space-separated tokens, order-independent, all optional:\n> - **scope** (`code` | `config` | `both`, default `both`) — which divergence dimension to run.\n> - **`repo=owner\u002Fname`** — override the parent repo whose forks are scanned (else auto-resolved).\n> - **`fork=owner\u002Fname`** — drill into a single fork (forces `code` scope; config math needs a fleet).\n>\n> Empty ⇒ **both** dimensions over the auto-resolved parent. Examples: `` (both, all forks) · `code` · `config` · `config repo=octo\u002Faeon` · `fork=alice\u002Faeon`.\n\nToday is ${today}. This is the fleet's **divergence monitor**. It answers two questions the popularity\u002Fliveness skills don't:\n1. **Code divergence** — which active forks are building real work (unique commits, new\u002Fmodified skills) that's worth pulling back upstream?\n2. **Config divergence** — where does the configured fleet systematically disagree with upstream's `enabled` \u002F `var` \u002F `model` \u002F `schedule` defaults, so the operator can flip a default the fleet has already voted on?\n\n`skill-gap` ranks **what's popular** (top 15 by enabled count). This skill's **code** branch surfaces **per-fork unique work**; its **config** branch surfaces **where operators disagree with defaults**. If 6 of 8 configured forks enable a skill upstream defaults off, upstream is shipping the wrong default; if 5 of 8 disable a skill upstream defaults on, that skill is noise. Both are peer-learning signals.\n\n## Operating principles\n\n- **Verdict first, catalog second.** The operator reads one line and knows if action is needed.\n- **Silent when nothing changed.** Weekly cadence + a dormant\u002Fundivergent fleet = a read-once habit to kill. A clean run notifies nothing.\n- **Per-fork compare is one call, not three.** `\u002Fcompare\u002F{owner}:main...{fork_owner}:main` returns ahead\u002Fbehind\u002Funique commits\u002Ffiles in a single round-trip; the recursive git-tree returns the fork's whole file list in one call.\n- **Substance ≠ noise.** A new `skills\u002F*\u002FSKILL.md` is worth 100 cron-time edits in `aeon.yml`. Score accordingly. On the config side, an untouched template fork is not a \"vote\" — exclude it from divergence math.\n\n---\n\n## Shared setup (all scopes)\n\n### S0. Bootstrap + load state\n\n**Shell shape.** Each Bash call is one allowed command (pipes are fine). The run refuses a call that sets a variable (`X=$(...)`, `X=\"${Y##*\u002F}\"`), nests `$(...)`, uses a `[ -f ... ] ||` test, or joins commands with `;`, `&&` or `||`, and each refusal costs a model turn. So every lookup below is its own call: read the value it prints and write it into later commands literally where they show `\u003CNAME>` (e.g. `\u003CPARENT_REPO>`, `\u003CFORK_FULL_NAME>`). Missing state files are found with Glob and created with Write; one call per fork, never a loop.\n\n```bash\nmkdir -p memory\u002Ftopics\n```\n\nCheck each of these with Glob; create any that does not exist with Write, with this content:\n- `memory\u002Finstances.json` - `{}`\n- `memory\u002Ftopics\u002Ffork-fleet-state.json` - `{\"forks\":{},\"last_run\":null}`\n- `memory\u002Ftopics\u002Ffork-digest-state.json` - `{\"last_run\":null}`\n\nRead `memory\u002FMEMORY.md` for high-level context and scan the last ~3 days of `memory\u002Flogs\u002F` — drop anything already reported so a weekly signal isn't re-sent.\n\n- Read `memory\u002Finstances.json` → the set of repo `full_name`s that are **managed instances** (tagged separately from organic community forks in the report).\n- Read `memory\u002Ftopics\u002Ffork-fleet-state.json` → prior run's per-fork `{pushed_at, ahead_by, default_branch, new_skill_count}` keyed by `full_name`. Used for the **code** what-changed delta.\n- Read `memory\u002Ftopics\u002Ffork-digest-state.json` → prior config-divergence snapshot (schema in step B8). Used for the **config** week-over-week delta.\n\n### S1. Parse the scope selector\n\nParse `${var}` into tokens:\n- `SCOPE` = `code`, `config`, or `both` (default `both` if no scope keyword present).\n- `REPO_OVERRIDE` = value of a `repo=owner\u002Fname` token, if any.\n- `SINGLE_FORK` = value of a `fork=owner\u002Fname` token, if any. **If `SINGLE_FORK` is set, force `SCOPE=code`** (single-fork config divergence is meaningless — the config math needs a fleet of ≥2 configured forks).\n\n### S2. Resolve the parent\u002Ftarget repo\n\nResolve `PARENT_REPO` in priority order:\n1. `REPO_OVERRIDE` if the `repo=` token was given.\n2. Else auto-resolve from this running instance, two calls:\n   ```bash\n   gh repo view --json nameWithOwner -q .nameWithOwner\n   ```\n   It prints SELF_REPO. Write it in literally:\n   ```bash\n   gh api repos\u002F\u003CSELF_REPO> --jq '.parent.full_name \u002F\u002F .full_name'\n   ```\n   It prints PARENT_REPO.\n3. Else fall back to the first non-comment, non-empty line of `memory\u002Fwatched-repos.md`.\n\nIf none resolves, write status `FORK_DIVERGENCE_NO_TARGET` to `memory\u002Flogs\u002F${today}.md` and stop (no notification).\n\n`PARENT_OWNER` = the part of `PARENT_REPO` before the `\u002F`, `PARENT_NAME` = the part after it (work these out yourself, no shell). Then, with PARENT_REPO written in literally:\n```bash\ngh api \"repos\u002F\u003CPARENT_REPO>\" --jq '.default_branch'\n```\nIt prints PARENT_DEFAULT_BRANCH.\n\n### S3. List + classify forks (single call, shared by both branches)\n\nOne paginated listing — includes `default_branch`, `archived`, `disabled`, `pushed_at`, stars, description:\n\n```bash\ngh api \"repos\u002F\u003CPARENT_REPO>\u002Fforks\" --paginate \\\n  --jq '[.[] | {full_name, owner: .owner.login, default_branch, pushed_at, pushed_at_epoch: (.pushed_at | fromdateiso8601), stargazers_count, open_issues_count, archived, disabled, description}]'\n```\n\nSkip `archived=true` or `disabled=true`. Retain the rest as the total fork population (`N_TOTAL`). Classify by activity window:\n\n- **Active** = `pushed_at` within last 30 days.\n- **Stale** = 30–365 days.\n- **Dormant** = >365 days or never pushed after creation.\n\nThe **Active** set (pushed in the last 30 days) is the shared working set for both branches — this matches the config branch's original 30-day cutoff exactly.\n\n- **If `SINGLE_FORK` is set:** filter to that one fork, treat it as active, and skip the classification math.\n- **If zero active forks** (and `SINGLE_FORK` unset): both branches short-circuit. If there is also no code-side state change (no new forks, none flipped active↔stale vs prior `fork-fleet-state.json`), write status `FORK_DIVERGENCE_QUIET` to the log, update the state files' `last_run`, send **no notification**, and stop.\n\nCap active-fork deep processing at **50 per run** — if more, rank by `pushed_at_epoch` desc and trim (log `truncated_at=50`).\n\nNow dispatch: run **Branch A** if `SCOPE ∈ {code, both}`, **Branch B** if `SCOPE ∈ {config, both}`.\n\n---\n\n## Branch A — Code divergence (runs when `SCOPE ∈ {code, both}`)\n\n### A1. Per-fork compare (one call each)\n\nFor each active fork, call cross-repo compare using the fork's own `default_branch` and `full_name` (absorbs any repo-rename drift), one call per fork with every value written in literally:\n\n```bash\ngh api \"repos\u002F\u003CPARENT_REPO>\u002Fcompare\u002F\u003CPARENT_OWNER>:\u003CPARENT_DEFAULT_BRANCH>...\u003CFORK_OWNER>:\u003CFORK_DEFAULT_BRANCH>\" \\\n  --jq '{ahead_by, behind_by, status, files: [.files[]? | {filename, status, additions, deletions}], commits: [.commits[]? | {sha: .sha[0:7], msg: .commit.message | split(\"\\n\")[0], author: .commit.author.name, date: .commit.author.date}]}'\n```\n\nOn `404` (branch missing \u002F fork emptied): mark fork `UNREADABLE` and continue.\nOn `429`: sleep 60s, retry once. On `5xx`: sleep 10s, retry once. On persistent fail: mark `API_FAIL` for that fork.\n\nCross-repo compare returns unique fork commits (`commits`) and changed files (up to 300) in one shot — no separate `\u002Fcommits` calls needed.\n\n### A2. Classify divergence signals per fork\n\nFrom the `files` array, tag each fork:\n- **New skills**: files with `status=added` under `skills\u002F*\u002FSKILL.md`\n- **Modified skills**: `status=modified` under `skills\u002F*\u002FSKILL.md`\n- **Custom schedule**: any change to `aeon.yml`\n- **Modified dashboard**: any change under `apps\u002Fdashboard\u002F`\n- **Custom notify**: change to `notify` or `notify-jsonrender`\n- **New content**: additions under `output\u002Farticles\u002F` or `memory\u002Ftopics\u002F`\n- **Config changes**: changes to `CLAUDE.md`, `.github\u002F`, `bin\u002F`, or root `scripts\u002F`\n- **Workflow changes**: changes under `.github\u002Fworkflows\u002F`\n\n### A3. Score each fork (substance-weighted)\n\n```\nscore =  10 × (new skill files)\n       +  4 × (modified skill files)\n       +  2 × min(unique_commits, 15)\n       +  3 × (new content files, capped at 5)\n       +  2 × (workflow\u002Fconfig files, capped at 3)\n       +  1 × (custom-schedule flag)\n       +  1 × stargazers\n```\n\nSort active forks by score descending. Flag any fork with ≥1 new skill file as a **PROMOTE** candidate; ≥3 unique commits OR ≥1 modified skill as **REVIEW**; otherwise **NOTE**.\n\n### A4. Deep-read top upstream candidates\n\nFor every PROMOTE fork (capped at 5), fetch each unique skill's SKILL.md from the fork's default branch:\n\n```bash\ngh api \"repos\u002F\u003CFORK_FULL_NAME>\u002Fcontents\u002F\u003CSKILL_PATH>?ref=\u003CFORK_DEFAULT_BRANCH>\" --jq '.content' | base64 -d\n```\n\nOn failure fall back to the file-tree listing and note \"could not read content\". Synthesize each unique skill into a 1–2 sentence description of what it does. Do **not** deep-read REVIEW or NOTE forks (output stays actionable).\n\n### A5. Compute week-over-week delta (code)\n\nCompare the current active-fork set to prior `fork-fleet-state.json`:\n- **NEW_FORK**: full_name absent from prior state\n- **NEW_ACTIVE**: was stale\u002Fdormant, now active\n- **WENT_STALE**: was active, now stale\u002Fdormant\n- **NEW_SKILLS**: active in both snapshots, `new_skill_count` increased\n- **GONE**: archived \u002F deleted since prior run\n\n### A6. Pick the code verdict\n\nOne line. Priority order:\n1. `NEW UPSTREAM CANDIDATE: {fork}` — if ≥1 PROMOTE fork has ≥1 new skill not present in prior state\n2. `ACTIVE FLEET: {N} forks building` — if ≥3 PROMOTE+REVIEW combined\n3. `FLEET STIRRING: {N} new active` — if ≥2 NEW_FORK or NEW_ACTIVE\n4. `HOLDING PATTERN: {N} active, no new work` — active forks present but nothing crossed REVIEW\n5. `DORMANT: no active forks` — shouldn't reach notify (S3 gates it); included for the log-only path\n\n### A7. Build the code-divergence article part\n\nAssemble this block (it becomes **Part 1** of the combined article in the final section):\n\n```markdown\n## What changed this week\n- **New forks**: [list or \"none\"]\n- **Went active**: [list or \"none\"]\n- **New skills landed**: [fork → skill names, or \"none\"]\n- **Went stale**: [list or \"none\"]\n- **Archived\u002Fdeleted**: [list or \"none\"]\n(Omit the entire section if every bucket is empty.)\n\n## PROMOTE - upstream contribution candidates\n\n### {fork_full_name} - score N [MANAGED | COMMUNITY]\n**Activity:** last pushed YYYY-MM-DD · stars N · +N\u002F-M commits vs upstream\n**Unique skills:**\n- `skills\u002Ffoo\u002FSKILL.md` - {one-line synthesis of what it does, from deep-read}\n- `skills\u002Fbar\u002FSKILL.md` - {synthesis}\n\n**Why promote:** {1-2 sentence take - what this skill does that upstream lacks, and whether it's generalizable}\n**Suggested action:** Open a PR cherry-picking `skills\u002Ffoo\u002F` (or reach out to {owner} to upstream themselves).\n\n(Repeat for each PROMOTE fork, capped at 5. If PROMOTE is empty: \"No upstream candidates this week.\")\n\n## REVIEW - worth a look\n\n| Fork | Score | Ahead | New\u002FModified | Notable |\n|------|-------|-------|--------------|---------|\n| owner\u002Frepo | N | +N\u002F-M | 0\u002F2 | dashboard rewrite, custom notify |\n\n(Omit if empty.)\n\n## NOTE - low divergence\n\nTerse one-liner per fork: `owner\u002Frepo (+N\u002F-M, schedule tweak only)`. Collapse into a count if >5 entries. Omit if empty.\n\n## Fleet vs community\n\n| Category | Count |\n|----------|-------|\n| Managed instances | N |\n| Community forks | N |\n| Stale (30-365d) | N |\n| Dormant (>365d) | N |\n\n## Code source status\n`forks_list=ok|fail · compare_ok=N\u002FM · deep_read=N\u002FM · rate_limit_retries=N · unreadable=N`\n```\n\nIf PROMOTE has >5 forks, keep only the top 5 by score; list the rest in REVIEW.\n\n### A8. Update code state\n\nWrite `memory\u002Ftopics\u002Ffork-fleet-state.json`:\n\n```json\n{\n  \"last_run\": \"${today}\",\n  \"last_status\": \"FORK_FLEET_OK\",\n  \"parent_repo\": \"owner\u002Frepo\",\n  \"forks\": {\n    \"owner\u002Frepo\": {\n      \"pushed_at\": \"YYYY-MM-DD...\",\n      \"default_branch\": \"main\",\n      \"ahead_by\": N,\n      \"behind_by\": N,\n      \"new_skill_count\": N,\n      \"score\": N,\n      \"tier\": \"PROMOTE|REVIEW|NOTE|UNREADABLE|API_FAIL\",\n      \"unique_skills\": [\"skills\u002Ffoo\u002FSKILL.md\", \"...\"]\n    }\n  }\n}\n```\n\n### A9. Set the code branch status\n\n| Status | Meaning |\n|--------|---------|\n| `FORK_FLEET_OK` | Active forks present AND (PROMOTE\u002FREVIEW non-empty OR delta non-empty) → contributes a notify signal |\n| `FORK_FLEET_NO_CHANGE` | Active forks exist but nothing crossed REVIEW and delta is empty → log only |\n| `FORK_FLEET_QUIET` | Zero active forks and no state change → log only |\n| `FORK_FLEET_API_FAIL` | Fork listing failed or >50% of compares failed → error signal |\n\n---\n\n## Branch B — Config divergence (runs when `SCOPE ∈ {config, both}`)\n\n### B1. Snapshot upstream defaults\n\nRead this running instance's local `aeon.yml` once. Build (these are baselines — never mutated):\n\n- `UPSTREAM_DEFAULTS`: dict `{skill_name -> {enabled: bool, model: str|null, var: str, schedule: str|null}}` for every skill entry under `skills:`.\n- `UPSTREAM_SKILLS`: set of skill directory names from `skills\u002F` (use `ls skills\u002F`).\n- `UPSTREAM_TAGS`: dict `{skill_name -> [tags]}` parsed from each `skills\u002F\u003Cname>\u002FSKILL.md` frontmatter (best-effort; missing frontmatter → `[]`).\n\n### B2. Per-fork enumeration (one tree call + one yml fetch each)\n\nOperate over the **active-fork set** from shared step S3 (already filtered to forks pushed in the last 30 days - the config branch's original cutoff). For each active fork, run **one** recursive git-tree call to enumerate files (cheaper than per-path contents), one call per fork with the fork written in literally:\n\n```bash\ngh api \"repos\u002F\u003CFORK_FULL>\u002Fgit\u002Ftrees\u002FHEAD?recursive=1\" --jq '[.tree[] | select(.type == \"blob\") | .path]'\n```\n\nThen fetch the fork's `aeon.yml` only if the tree contains it:\n\n```bash\ngh api \"repos\u002F\u003CFORK_FULL>\u002Fcontents\u002Faeon.yml?ref=\u003CFORK_DEFAULT_BRANCH>\" --jq '.content' | base64 -d\n```\n\nError handling:\n- 404 \u002F 409 (empty repo): mark `status: \"no_tree\"`, skip aeon.yml extraction, continue.\n- 403 with `X-RateLimit-Remaining: 0`: sleep 60s, retry once. If still failing, mark `status: \"rate_limited\"` and continue.\n- Tree contains aeon.yml but the contents call 404s: mark `status: \"yml_unreadable\"`, continue.\n- aeon.yml present but YAML parse fails: mark `status: \"yml_invalid\"`, continue.\n\nFor each readable `aeon.yml`, extract per-skill `{enabled, model, var, schedule}`. Treat missing keys as inheriting the upstream default (do **not** count those as overrides).\n\nDetect **fork-only skills**: directory names matching `skills\u002F\u003Cname>\u002FSKILL.md` in the fork's tree where `\u003Cname>` is NOT in `UPSTREAM_SKILLS`. Record `{fork_full_name, skill_name, path}` for each.\n\n### B3. Tier each fork\n\nCompute a divergence signal vector vs `UPSTREAM_DEFAULTS`:\n- `enabled_diff`: count of skills where the fork's `enabled` differs from upstream\n- `var_overrides`: count of skills with non-empty `var:` where upstream's was empty (or a different non-empty value)\n- `model_overrides`: count of skills with `model:` differing from upstream\n- `schedule_overrides`: count of skills with `schedule:` differing from upstream\n- `fork_only_skill_count`: count from step B2\n\nTier the fork:\n- **CONFIGURED**: any signal ≥1 (the fork actively diverged)\n- **TEMPLATE**: aeon.yml readable but every signal is 0 — excluded from divergence math\n- **UNREADABLE**: no_tree \u002F no aeon.yml \u002F yml_unreadable \u002F yml_invalid \u002F rate_limited — tracked in the source-status footer\n\nLet `N_CONFIGURED` = count of forks tiered CONFIGURED. **If `N_CONFIGURED \u003C 2`:** the config branch cannot produce meaningful divergence math. Set config status `FORK_SKILL_DIGEST_TEMPLATE_FLEET`, record active\u002Ftemplate\u002Funreadable counts, emit a stub Part 2 noting the conversion rate, and contribute **no** config notify signal. Skip steps B4–B6.\n\n### B4. Aggregate divergence (the core config analysis)\n\nFor each skill name in `UPSTREAM_SKILLS`, compute four dimensions:\n\n**Enable divergence:**\n- `forks_enabled_count`: number of CONFIGURED forks with `enabled: true` for this skill\n- `forks_disabled_count`: number of CONFIGURED forks with `enabled: false` (explicitly set, not inherited)\n- `upstream_enabled`: bool from UPSTREAM_DEFAULTS\n- `divergence_pct`:\n  - If upstream `enabled: false`: `forks_enabled_count \u002F N_CONFIGURED` (how many disagree by enabling)\n  - If upstream `enabled: true`: `forks_disabled_count \u002F N_CONFIGURED` (how many disagree by disabling)\n- `direction`: `\"ENABLE_UPWARD\"` (upstream off, forks turn on) or `\"DISABLE_DOWNWARD\"` (upstream on, forks turn off)\n\n**Var divergence:**\n- `var_override_count`: number of CONFIGURED forks where `var:` differs from upstream\n- `top_var_value`: most common non-empty fork value (with count) — only if ≥2 forks share it\n\n**Model divergence:**\n- `model_override_count`: number of forks with non-null model differing from upstream\n- `top_model_value`: most common fork model (with count) — only if ≥2 forks share it (signals fleet consensus on a cheaper\u002Fdifferent model)\n\n**Schedule divergence:**\n- `schedule_override_count`: number of forks with schedule differing from upstream\n- `top_schedule_value`: most common fork schedule (with count) — only if ≥2 forks share it\n\n### B5. Categorize divergent skills\n\nClassify each skill into **at most one** bucket (first match wins, in this order):\n\n- **DEFAULT_FLIP_ENABLE**: `direction == \"ENABLE_UPWARD\"` AND `divergence_pct >= 0.50` AND skill is not `workflow_dispatch` AND skill not tagged `meta`\u002F`dev`. Recommend: flip upstream default to `enabled: true`.\n- **DEFAULT_FLIP_DISABLE**: `direction == \"DISABLE_DOWNWARD\"` AND `divergence_pct >= 0.50`. Recommend: flip upstream default to `enabled: false` (the fleet is voting it as noise).\n- **MODEL_CONSENSUS**: `top_model_value` non-null AND its count `>= max(2, ceil(N_CONFIGURED * 0.40))`. Recommend: match the fleet's model in upstream.\n- **VAR_HOTSPOT**: `var_override_count >= max(2, ceil(N_CONFIGURED * 0.30))` AND `top_var_value` non-null. Recommend: surface the common var value in upstream docs or as the default.\n- **EMERGING**: `direction == \"ENABLE_UPWARD\"` AND `0.25 \u003C= divergence_pct \u003C 0.50` AND not already in a flip bucket. Surface as a watchlist — fleet sentiment building but not yet majority.\n- (otherwise: not categorized; appears only in the appendix divergence table if any signal is non-zero)\n\nSkills with all-zero divergence are omitted.\n\n### B6. Per-fork customization fingerprint\n\nFor each CONFIGURED fork:\n- `total_overrides`: `enabled_diff + var_overrides + model_overrides + schedule_overrides + fork_only_skill_count`\n- `category_lean`: dict `{tag -> count_of_enabled_skills_with_that_tag}` (using UPSTREAM_TAGS for upstream skills the fork enables; fork-only skills counted under tag `\"fork-only\"`)\n- `dominant_category`: tag with max count, or `\"mixed\"` if no tag holds >40% of total enabled count\n\nRank forks by `total_overrides` desc. Top 5 = \"heaviest customizers\" - surface with dominant category and a one-line synthesis (e.g. `\"owner\u002Faeon - content-heavy: 14 article\u002Fdigest skills enabled, 3 model overrides to claude-sonnet-5-5\"`). The fingerprint is **descriptive only** - never recommend changes to individual forks.\n\n### B7. Config week-over-week delta\n\nRead the prior `memory\u002Ftopics\u002Ffork-digest-state.json` snapshot (schema in B8). If it exists and `last_run` is within the last 14 days, compute:\n- **NEW_FLIP**: skills now in DEFAULT_FLIP_* that weren't last run\n- **STRENGTHENED**: skills that moved EMERGING → DEFAULT_FLIP_ENABLE\n- **FADED**: skills that left a flip bucket since last run\n- **NEW_FORK_ONLY**: fork-only skills not present last run\n- **NEW_HEAVY_CUSTOMIZER**: forks now in the top-5 fingerprint that weren't before\n\nIf the file is missing or stale (>14 days), set all deltas to `\"first divergence snapshot\"`.\n\n### B8. Pick the config verdict + persist snapshot\n\nConfig verdict line, strongest single claim first:\n1. Any `DEFAULT_FLIP_ENABLE`: `\"${N} forks enable ${skill} (upstream defaults off) — flip the default\"`\n2. Else any `DEFAULT_FLIP_DISABLE`: `\"${N} forks disable ${skill} (upstream defaults on) — fleet is voting it as noise\"`\n3. Else any `MODEL_CONSENSUS`: `\"${N} forks override ${skill} → ${model} — match upstream\"`\n4. Else any `NEW_FORK_ONLY` from delta: `\"${fork_owner} shipped ${skill} — not in upstream\"`\n5. Else any `EMERGING`: `\"${skill} adoption building (${pct}% of configured) — watchlist\"`\n6. Else: `\"${N_CONFIGURED} configured forks; no divergence pattern crossed flip threshold\"`\n\nPersist `memory\u002Ftopics\u002Ffork-digest-state.json` (overwrite each run — the JSON is the delta contract; do NOT parse last week's article):\n\n```json\n{\n  \"last_run\": \"${today}\",\n  \"target_repo\": \"${PARENT_REPO}\",\n  \"n_active\": N_ACTIVE,\n  \"n_configured\": N_CONFIGURED,\n  \"n_template\": N_TEMPLATE,\n  \"n_unreadable\": N_UNREADABLE,\n  \"buckets\": {\n    \"DEFAULT_FLIP_ENABLE\": [{\"skill\": \"name\", \"forks\": N, \"pct\": 0.NN}],\n    \"DEFAULT_FLIP_DISABLE\": [{\"skill\": \"name\", \"forks\": N, \"pct\": 0.NN}],\n    \"MODEL_CONSENSUS\": [{\"skill\": \"name\", \"model\": \"value\", \"forks\": N}],\n    \"VAR_HOTSPOT\": [{\"skill\": \"name\", \"var\": \"value\", \"forks\": N}],\n    \"EMERGING\": [{\"skill\": \"name\", \"pct\": 0.NN}]\n  },\n  \"fork_only_skills\": [{\"fork\": \"owner\u002Frepo\", \"skill\": \"name\"}],\n  \"fingerprints\": [{\"fork\": \"owner\u002Frepo\", \"total_overrides\": N, \"dominant_category\": \"tag\"}]\n}\n```\n\n### B9. Build the config-divergence article part\n\nAssemble this block (it becomes **Part 2** of the combined article):\n\n```markdown\n*Scanned ${N_ACTIVE} active forks of ${PARENT_REPO} (pushed in last 30 days). ${N_CONFIGURED} are configured (aeon.yml diverges from upstream defaults). Divergence scored against the configured ${N_CONFIGURED}.*\n\n## Default-flip candidates\n\n### Enable upward (upstream off → fleet enables)\n| Skill | Forks enabled | % of configured | Δ vs last week |\n|-------|---------------|-----------------|----------------|\n| name  | N             | XX%             | NEW \u002F STRENGTHENED \u002F - |\n\n(Only DEFAULT_FLIP_ENABLE. If empty: \"No skills crossed the 50% enable-upward threshold this week.\")\n\n### Disable downward (upstream on → fleet disables)\n| Skill | Forks disabled | % of configured | Δ vs last week |\n|-------|----------------|-----------------|----------------|\n| name  | N              | XX%             | NEW \u002F - |\n\n(Only DEFAULT_FLIP_DISABLE. If empty: \"No skills crossed the 50% disable-downward threshold.\")\n\n## Fleet consensus on alternative settings\n\n### Model overrides\n${MODEL_CONSENSUS entries: \"skill X - N forks → claude-sonnet-5-5 (40% of configured)\" OR \"none this week\"}\n\n### Var hotspots\n${VAR_HOTSPOT entries: \"skill X - N forks set var to '${value}'\" OR \"none this week\"}\n\n### Schedule overrides\n${skills where ≥2 forks share an alternative schedule, with the schedule string OR \"none this week\"}\n\n## Watchlist (emerging - 25-49% adoption)\n${EMERGING skills with adoption % OR \"none this week\"}\n\n## Heaviest customizers (top 5)\n\n| Fork | Total overrides | Dominant category | Notes |\n|------|-----------------|-------------------|-------|\n| owner\u002Frepo | N | content \u002F dev \u002F meta \u002F fork-only \u002F mixed | one-line synthesis |\n\n## Fork-only skills\n\n${list of {fork, skill_name} pairs OR \"none this week\"}\n\n(These skills exist as `skills\u002F\u003Cname>\u002FSKILL.md` in a fork but not in upstream - fork experiments worth reviewing for upstreaming.)\n\n## Config week-over-week\n\n${\"First divergence snapshot - no comparison\" OR list of NEW_FLIP \u002F STRENGTHENED \u002F FADED \u002F NEW_FORK_ONLY \u002F NEW_HEAVY_CUSTOMIZER}\n\n## Fleet composition (config tiers)\n\n| Tier | Count | % |\n|------|-------|---|\n| Configured | N_CONFIGURED | XX% |\n| Template (untouched aeon.yml) | N_TEMPLATE | XX% |\n| Unreadable | N_UNREADABLE | XX% |\n| **Total active** | N_ACTIVE | 100% |\n\n## Config source status\n- Trees fetched: N_TREES_OK \u002F N_ACTIVE\n- aeon.yml readable: (N_CONFIGURED + N_TEMPLATE) \u002F N_ACTIVE\n- YAML parse failures: N_YML_INVALID\n- Rate-limited: N_RATE_LIMITED\n- Fork-only skills inspected: N_FORK_ONLY_FILES\n\n## Appendix - full divergence table\n\n(Every skill with ≥1 non-zero divergence signal, sorted by total override count desc. Columns: skill, enable_diff, var_overrides, model_overrides, schedule_overrides. Cap at 30 rows; if more, append \"+ N more skills with low-signal divergence\".)\n```\n\n### Config branch status\n\n| Status | Meaning |\n|--------|---------|\n| `FORK_SKILL_DIGEST_OK` | ≥2 configured forks AND ≥1 flip\u002Fconsensus\u002Fnew-fork-only signal → contributes a notify signal |\n| `FORK_SKILL_DIGEST_QUIET` | ≥2 configured forks but no signal crossed thresholds → log only |\n| `FORK_SKILL_DIGEST_TEMPLATE_FLEET` | \u003C2 configured forks (mostly templates) → log only |\n| `FORK_SKILL_DIGEST_NO_FORKS` | Zero active forks → log only |\n\n---\n\n## Assemble the report (all scopes)\n\n### R1. Write the combined article\n\nTo `output\u002Farticles\u002Ffork-divergence-${today}.md`. Header first, then whichever parts ran:\n\n```markdown\n# Fork Divergence - ${today}\n\n**Verdict:** {lead with the stronger of the two sub-verdicts - a code PROMOTE\u002FNEW-UPSTREAM-CANDIDATE outranks a config flip only if it's a genuinely new skill; otherwise a DEFAULT_FLIP leads. Use judgment; one line.}\n\n- **Code divergence:** {code verdict from A6, or \"not run (scope=config)\"}\n- **Config divergence:** {config verdict from B8, or \"not run (scope=code)\"}\n\nFleet: N_TOTAL total forks · N_ACTIVE active · N_MANAGED managed instances · N_COMMUNITY community.\n\n---\n\n# Part 1 - Code divergence\n{Branch A article part from A7; omit this whole part if SCOPE=config}\n\n---\n\n# Part 2 - Config divergence\n{Branch B article part from B9; omit this whole part if SCOPE=code}\n\n---\n*Source: GitHub API - forks of ${PARENT_REPO}. Code divergence = per-fork unique commits\u002Fskills vs upstream. Config divergence = where configured forks' aeon.yml disagrees with upstream `enabled`\u002F`var`\u002F`model`\u002F`schedule`; untouched templates are excluded from the config math. Companion to `skill-gap` (popularity).*\n```\n\nCap the article at ~700 lines total (≈500 for code sections when both parts run). When only one branch ran, drop the other Part heading and the missing sub-verdict line.\n\n### R2. Notify — gated\n\nRead `soul\u002F` (if present) to match the operator's voice. **Skip notify entirely** when neither branch produced a signal, i.e. when:\n- code status ∈ {`FORK_FLEET_NO_CHANGE`, `FORK_FLEET_QUIET`} (or code didn't run), **AND**\n- config status ∈ {`FORK_SKILL_DIGEST_QUIET`, `FORK_SKILL_DIGEST_TEMPLATE_FLEET`, `FORK_SKILL_DIGEST_NO_FORKS`} (or config didn't run).\n\nIf either branch hit `FORK_FLEET_API_FAIL`, send an **error** notify (`--severity warn`) noting the failure and source status.\n\nOtherwise send one combined message via `.\u002Fnotify` (include only the sub-blocks whose branch produced signal; keep it tight):\n\n```\n*Fork Divergence - ${today}*\n{combined verdict line}\n\nFleet: N_ACTIVE active \u002F N_TOTAL total. {1 sentence on shape - \"mostly managed instances\", \"community picking up\", \"template-heavy\", etc.}\n\n{If code PROMOTE non-empty:}\nUpstream candidate: {top PROMOTE fork}\n{2 sentences: what they built, why it's worth merging back}\n\n{If code delta has NEW_SKILLS:}\nNew skills landed this week:\n- {fork} → `skills\u002Ffoo\u002FSKILL.md` - {synthesis}\n\n{If DEFAULT_FLIP_ENABLE non-empty (top 3):}\nFlip enable (upstream off → fleet on):\n- {skill} - {N} forks ({pct}%)\n\n{If DEFAULT_FLIP_DISABLE non-empty (top 3):}\nFlip disable (upstream on → fleet off):\n- {skill} - {N} forks ({pct}%)\n\n{If MODEL_CONSENSUS non-empty (top 2):}\nModel consensus:\n- {skill} → {model} ({N} forks)\n\n{If config delta NEW_FORK_ONLY non-empty:}\nNew fork-only skills: {comma-separated owner\u002Fskill, capped at 3}\n\nFull report: https:\u002F\u002Fgithub.com\u002F\u003CSELF_REPO>\u002Fblob\u002Fmain\u002Foutput\u002Farticles\u002Ffork-divergence-${today}.md\n```\n\nUse this running instance's repo (`GITHUB_REPOSITORY`) for the URL (the article lives in this running instance's repo, not the target repo). It is SELF_REPO, printed by `gh repo view --json nameWithOwner -q .nameWithOwner` (run that as its own call if S2 did not already); write it in literally.\n\n### R3. Log\n\nAppend to `memory\u002Flogs\u002F${today}.md` under **one** heading. Include a discriminator line naming the scope that ran, then only the sub-blocks for branches that ran:\n\n```\n### fork-fleet\n- Scope: {both | code | config}  ·  Combined status: {FORK_DIVERGENCE_OK | NO_CHANGE | QUIET | NO_TARGET | API_FAIL}\n- Verdict: {combined verdict line}\n- Fleet: N_ACTIVE active \u002F N_TOTAL total (N_MANAGED managed, N_COMMUNITY community)\n\n[code]  (only if Branch A ran)\n- Code status: {FORK_FLEET_OK | NO_CHANGE | QUIET | API_FAIL}\n- PROMOTE: N forks (list), REVIEW: N, NOTE: N\n- Code delta: {new_forks:N, new_active:N, new_skills:N, went_stale:N}\n- Code source: forks_list=ok|fail · compare_ok=N\u002FM · deep_read=N\u002FM · unreadable=N\n\n[config]  (only if Branch B ran)\n- Config status: {FORK_SKILL_DIGEST_OK | QUIET | TEMPLATE_FLEET | NO_FORKS}\n- Configured: N (XX% conversion) · Template: N · Unreadable: N\n- DEFAULT_FLIP_ENABLE: N · DEFAULT_FLIP_DISABLE: N · MODEL_CONSENSUS: N · VAR_HOTSPOT: N · EMERGING: N\n- Fork-only skills: N · Heaviest customizer: {fork} ({N} overrides)\n\n- Article: output\u002Farticles\u002Ffork-divergence-${today}.md\n- Notification sent: yes\u002Fno\n```\n\n## Exit taxonomy (combined)\n\nThe combined status rolls up the per-branch statuses (kept verbatim in A9 \u002F B-status above):\n\n| Combined status | Rolls up when | Notify? |\n|-----------------|---------------|---------|\n| `FORK_DIVERGENCE_OK` | code = `FORK_FLEET_OK` **OR** config = `FORK_SKILL_DIGEST_OK` | Yes |\n| `FORK_DIVERGENCE_NO_CHANGE` | both branches ran but neither reached OK (all `NO_CHANGE`\u002F`QUIET`\u002F`TEMPLATE_FLEET`\u002F`NO_FORKS`) | No (log only) |\n| `FORK_DIVERGENCE_QUIET` | zero active forks and no code-side state change (S3 short-circuit) | No (log only) |\n| `FORK_DIVERGENCE_NO_TARGET` | no parent\u002Ftarget repo resolved (S2) | No (log only) |\n| `FORK_DIVERGENCE_API_FAIL` | fork listing failed, or >50% of a branch's compares\u002Ftrees failed | Yes (error notify) |\n\n## Constraints\n\n- **Cross-repo compare** accepts up to 300 files per response; if a fork exceeds this, note `files_truncated=true` for it and proceed.\n- Cap active-fork deep processing at **50 per run** (S3) — rank by `pushed_at_epoch` desc and trim, logging `truncated_at=50`.\n- Never deep-read content from a fork with `archived=true`, or when the SKILL.md path is absent from the compare `files` list (cheapest sanity check).\n- **Never invent a PROMOTE candidate** — a fork with zero new skill files is at most REVIEW.\n- **Config math needs a denominator:** never send a config signal when `N_CONFIGURED \u003C 2` — the divergence percentages are meaningless without a configured base.\n- Skills tagged `meta` or `dev` are excluded from `DEFAULT_FLIP_ENABLE` (operator tools — fork adoption isn't the success metric). They can still appear in MODEL_CONSENSUS, VAR_HOTSPOT, and the appendix.\n- Skills with `schedule: \"workflow_dispatch\"` are excluded from **both** flip buckets (on-demand by design — adoption % is misleading).\n- `heartbeat` is excluded from `DEFAULT_FLIP_DISABLE` (every fork inheriting upstream's `enabled: true` would game the disable count if any fork explicitly disables it to stay quiet).\n- The per-fork fingerprint is descriptive only — only aggregate signals drive recommendations.\n- Silent runs are **correct**, not failures. This skill is the divergence companion to `skill-gap` (popularity); avoid duplicating its headline metrics — focus on the code + config **divergence patterns** it doesn't surface.\n\n## Network note\n\nEvery GitHub call uses `gh api`, which authenticates via `GITHUB_TOKEN` automatically — no `curl`, no `$SECRET` on the command line (so nothing for the Bash permission layer to refuse), no secrets beyond the default `GITHUB_TOKEN`. Retry policy: on `429`\u002F`5xx` (compare) back off per step A1; on `403` with `X-RateLimit-Remaining: 0` (tree\u002Fcontents) sleep 60s and retry once, then mark that fork `rate_limited` and proceed with a partial fleet (the verdict and source-status footers surface the gap). If the initial `\u002Fforks` listing fails after retry, combined status = `FORK_DIVERGENCE_API_FAIL` with `forks_list=fail`.\n","> **${var}** — Divergence scope selector; space-separated tokens, order-independent, all optional:\n> - **scope** (`code` | `config` | `both`, default `both`) — which divergence dimension to run.\n> - **`repo=owner\u002Fname`** — override the parent repo whose forks are scanned (else auto-resolved).\n> - **`fork=owner\u002Fname`** — drill into a single fork (forces `code` scope; config math needs a fleet).\n>\n> Empty ⇒ **both** dimensions over the auto-resolved parent. Examples: `` (both, all forks) · `code` · `config` · `config repo=octo\u002Faeon` · `fork=alice\u002Faeon`.\n\nToday is ${today}. This is the fleet's **divergence monitor**. It answers two questions the popularity\u002Fliveness skills don't:\n1. **Code divergence** — which active forks are building real work (unique commits, new\u002Fmodified skills) that's worth pulling back upstream?\n2. **Config divergence** — where does the configured fleet systematically disagree with upstream's `enabled` \u002F `var` \u002F `model` \u002F `schedule` defaults, so the operator can flip a default the fleet has already voted on?\n\n`skill-gap` ranks **what's popular** (top 15 by enabled count). This skill's **code** branch surfaces **per-fork unique work**; its **config** branch surfaces **where operators disagree with defaults**. If 6 of 8 configured forks enable a skill upstream defaults off, upstream is shipping the wrong default; if 5 of 8 disable a skill upstream defaults on, that skill is noise. Both are peer-learning signals.\n\n## Operating principles\n\n- **Verdict first, catalog second.** The operator reads one line and knows if action is needed.\n- **Silent when nothing changed.** Weekly cadence + a dormant\u002Fundivergent fleet = a read-once habit to kill. A clean run notifies nothing.\n- **Per-fork compare is one call, not three.** `\u002Fcompare\u002F{owner}:main...{fork_owner}:main` returns ahead\u002Fbehind\u002Funique commits\u002Ffiles in a single round-trip; the recursive git-tree returns the fork's whole file list in one call.\n- **Substance ≠ noise.** A new `skills\u002F*\u002FSKILL.md` is worth 100 cron-time edits in `aeon.yml`. Score accordingly. On the config side, an untouched template fork is not a \"vote\" — exclude it from divergence math.\n\n---\n\n## Shared setup (all scopes)\n\n### S0. Bootstrap + load state\n\n**Shell shape.** Each Bash call is one allowed command (pipes are fine). The run refuses a call that sets a variable (`X=$(...)`, `X=\"${Y##*\u002F}\"`), nests `$(...)`, uses a `[ -f ... ] ||` test, or joins commands with `;`, `&&` or `||`, and each refusal costs a model turn. So every lookup below is its own call: read the value it prints and write it into later commands literally where they show `\u003CNAME>` (e.g. `\u003CPARENT_REPO>`, `\u003CFORK_FULL_NAME>`). Missing state files are found with Glob and created with Write; one call per fork, never a loop.\n\n```bash\nmkdir -p memory\u002Ftopics\n```\n\nCheck each of these with Glob; create any that does not exist with Write, with this content:\n- `memory\u002Finstances.json` - `{}`\n- `memory\u002Ftopics\u002Ffork-fleet-state.json` - `{\"forks\":{},\"last_run\":null}`\n- `memory\u002Ftopics\u002Ffork-digest-state.json` - `{\"last_run\":null}`\n\nRead `memory\u002FMEMORY.md` for high-level context and scan the last ~3 days of `memory\u002Flogs\u002F` — drop anything already reported so a weekly signal isn't re-sent.\n\n- Read `memory\u002Finstances.json` → the set of repo `full_name`s that are **managed instances** (tagged separately from organic community forks in the report).\n- Read `memory\u002Ftopics\u002Ffork-fleet-state.json` → prior run's per-fork `{pushed_at, ahead_by, default_branch, new_skill_count}` keyed by `full_name`. Used for the **code** what-changed delta.\n- Read `memory\u002Ftopics\u002Ffork-digest-state.json` → prior config-divergence snapshot (schema in step B8). Used for the **config** week-over-week delta.\n\n### S1. Parse the scope selector\n\nParse `${var}` into tokens:\n- `SCOPE` = `code`, `config`, or `both` (default `both` if no scope keyword present).\n- `REPO_OVERRIDE` = value of a `repo=owner\u002Fname` token, if any.\n- `SINGLE_FORK` = value of a `fork=owner\u002Fname` token, if any. **If `SINGLE_FORK` is set, force `SCOPE=code`** (single-fork config divergence is meaningless — the config math needs a fleet of ≥2 configured forks).\n\n### S2. Resolve the parent\u002Ftarget repo\n\nResolve `PARENT_REPO` in priority order:\n1. `REPO_OVERRIDE` if the `repo=` token was given.\n2. Else auto-resolve from this running instance, two calls:\n   ```bash\n   gh repo view --json nameWithOwner -q .nameWithOwner\n   ```\n   It prints SELF_REPO. Write it in literally:\n   ```bash\n   gh api repos\u002F\u003CSELF_REPO> --jq '.parent.full_name \u002F\u002F .full_name'\n   ```\n   It prints PARENT_REPO.\n3. Else fall back to the first non-comment, non-empty line of `memory\u002Fwatched-repos.md`.\n\nIf none resolves, write status `FORK_DIVERGENCE_NO_TARGET` to `memory\u002Flogs\u002F${today}.md` and stop (no notification).\n\n`PARENT_OWNER` = the part of `PARENT_REPO` before the `\u002F`, `PARENT_NAME` = the part after it (work these out yourself, no shell). Then, with PARENT_REPO written in literally:\n```bash\ngh api \"repos\u002F\u003CPARENT_REPO>\" --jq '.default_branch'\n```\nIt prints PARENT_DEFAULT_BRANCH.\n\n### S3. List + classify forks (single call, shared by both branches)\n\nOne paginated listing — includes `default_branch`, `archived`, `disabled`, `pushed_at`, stars, description:\n\n```bash\ngh api \"repos\u002F\u003CPARENT_REPO>\u002Fforks\" --paginate \\\n  --jq '[.[] | {full_name, owner: .owner.login, default_branch, pushed_at, pushed_at_epoch: (.pushed_at | fromdateiso8601), stargazers_count, open_issues_count, archived, disabled, description}]'\n```\n\nSkip `archived=true` or `disabled=true`. Retain the rest as the total fork population (`N_TOTAL`). Classify by activity window:\n\n- **Active** = `pushed_at` within last 30 days.\n- **Stale** = 30–365 days.\n- **Dormant** = >365 days or never pushed after creation.\n\nThe **Active** set (pushed in the last 30 days) is the shared working set for both branches — this matches the config branch's original 30-day cutoff exactly.\n\n- **If `SINGLE_FORK` is set:** filter to that one fork, treat it as active, and skip the classification math.\n- **If zero active forks** (and `SINGLE_FORK` unset): both branches short-circuit. If there is also no code-side state change (no new forks, none flipped active↔stale vs prior `fork-fleet-state.json`), write status `FORK_DIVERGENCE_QUIET` to the log, update the state files' `last_run`, send **no notification**, and stop.\n\nCap active-fork deep processing at **50 per run** — if more, rank by `pushed_at_epoch` desc and trim (log `truncated_at=50`).\n\nNow dispatch: run **Branch A** if `SCOPE ∈ {code, both}`, **Branch B** if `SCOPE ∈ {config, both}`.\n\n---\n\n## Branch A — Code divergence (runs when `SCOPE ∈ {code, both}`)\n\n### A1. Per-fork compare (one call each)\n\nFor each active fork, call cross-repo compare using the fork's own `default_branch` and `full_name` (absorbs any repo-rename drift), one call per fork with every value written in literally:\n\n```bash\ngh api \"repos\u002F\u003CPARENT_REPO>\u002Fcompare\u002F\u003CPARENT_OWNER>:\u003CPARENT_DEFAULT_BRANCH>...\u003CFORK_OWNER>:\u003CFORK_DEFAULT_BRANCH>\" \\\n  --jq '{ahead_by, behind_by, status, files: [.files[]? | {filename, status, additions, deletions}], commits: [.commits[]? | {sha: .sha[0:7], msg: .commit.message | split(\"\\n\")[0], author: .commit.author.name, date: .commit.author.date}]}'\n```\n\nOn `404` (branch missing \u002F fork emptied): mark fork `UNREADABLE` and continue.\nOn `429`: sleep 60s, retry once. On `5xx`: sleep 10s, retry once. On persistent fail: mark `API_FAIL` for that fork.\n\nCross-repo compare returns unique fork commits (`commits`) and changed files (up to 300) in one shot — no separate `\u002Fcommits` calls needed.\n\n### A2. Classify divergence signals per fork\n\nFrom the `files` array, tag each fork:\n- **New skills**: files with `status=added` under `skills\u002F*\u002FSKILL.md`\n- **Modified skills**: `status=modified` under `skills\u002F*\u002FSKILL.md`\n- **Custom schedule**: any change to `aeon.yml`\n- **Modified dashboard**: any change under `apps\u002Fdashboard\u002F`\n- **Custom notify**: change to `notify` or `notify-jsonrender`\n- **New content**: additions under `output\u002Farticles\u002F` or `memory\u002Ftopics\u002F`\n- **Config changes**: changes to `CLAUDE.md`, `.github\u002F`, `bin\u002F`, or root `scripts\u002F`\n- **Workflow changes**: changes under `.github\u002Fworkflows\u002F`\n\n### A3. Score each fork (substance-weighted)\n\n```\nscore =  10 × (new skill files)\n       +  4 × (modified skill files)\n       +  2 × min(unique_commits, 15)\n       +  3 × (new content files, capped at 5)\n       +  2 × (workflow\u002Fconfig files, capped at 3)\n       +  1 × (custom-schedule flag)\n       +  1 × stargazers\n```\n\nSort active forks by score descending. Flag any fork with ≥1 new skill file as a **PROMOTE** candidate; ≥3 unique commits OR ≥1 modified skill as **REVIEW**; otherwise **NOTE**.\n\n### A4. Deep-read top upstream candidates\n\nFor every PROMOTE fork (capped at 5), fetch each unique skill's SKILL.md from the fork's default branch:\n\n```bash\ngh api \"repos\u002F\u003CFORK_FULL_NAME>\u002Fcontents\u002F\u003CSKILL_PATH>?ref=\u003CFORK_DEFAULT_BRANCH>\" --jq '.content' | base64 -d\n```\n\nOn failure fall back to the file-tree listing and note \"could not read content\". Synthesize each unique skill into a 1–2 sentence description of what it does. Do **not** deep-read REVIEW or NOTE forks (output stays actionable).\n\n### A5. Compute week-over-week delta (code)\n\nCompare the current active-fork set to prior `fork-fleet-state.json`:\n- **NEW_FORK**: full_name absent from prior state\n- **NEW_ACTIVE**: was stale\u002Fdormant, now active\n- **WENT_STALE**: was active, now stale\u002Fdormant\n- **NEW_SKILLS**: active in both snapshots, `new_skill_count` increased\n- **GONE**: archived \u002F deleted since prior run\n\n### A6. Pick the code verdict\n\nOne line. Priority order:\n1. `NEW UPSTREAM CANDIDATE: {fork}` — if ≥1 PROMOTE fork has ≥1 new skill not present in prior state\n2. `ACTIVE FLEET: {N} forks building` — if ≥3 PROMOTE+REVIEW combined\n3. `FLEET STIRRING: {N} new active` — if ≥2 NEW_FORK or NEW_ACTIVE\n4. `HOLDING PATTERN: {N} active, no new work` — active forks present but nothing crossed REVIEW\n5. `DORMANT: no active forks` — shouldn't reach notify (S3 gates it); included for the log-only path\n\n### A7. Build the code-divergence article part\n\nAssemble this block (it becomes **Part 1** of the combined article in the final section):\n\n```markdown\n## What changed this week\n- **New forks**: [list or \"none\"]\n- **Went active**: [list or \"none\"]\n- **New skills landed**: [fork → skill names, or \"none\"]\n- **Went stale**: [list or \"none\"]\n- **Archived\u002Fdeleted**: [list or \"none\"]\n(Omit the entire section if every bucket is empty.)\n\n## PROMOTE - upstream contribution candidates\n\n### {fork_full_name} - score N [MANAGED | COMMUNITY]\n**Activity:** last pushed YYYY-MM-DD · stars N · +N\u002F-M commits vs upstream\n**Unique skills:**\n- `skills\u002Ffoo\u002FSKILL.md` - {one-line synthesis of what it does, from deep-read}\n- `skills\u002Fbar\u002FSKILL.md` - {synthesis}\n\n**Why promote:** {1-2 sentence take - what this skill does that upstream lacks, and whether it's generalizable}\n**Suggested action:** Open a PR cherry-picking `skills\u002Ffoo\u002F` (or reach out to {owner} to upstream themselves).\n\n(Repeat for each PROMOTE fork, capped at 5. If PROMOTE is empty: \"No upstream candidates this week.\")\n\n## REVIEW - worth a look\n\n| Fork | Score | Ahead | New\u002FModified | Notable |\n|------|-------|-------|--------------|---------|\n| owner\u002Frepo | N | +N\u002F-M | 0\u002F2 | dashboard rewrite, custom notify |\n\n(Omit if empty.)\n\n## NOTE - low divergence\n\nTerse one-liner per fork: `owner\u002Frepo (+N\u002F-M, schedule tweak only)`. Collapse into a count if >5 entries. Omit if empty.\n\n## Fleet vs community\n\n| Category | Count |\n|----------|-------|\n| Managed instances | N |\n| Community forks | N |\n| Stale (30-365d) | N |\n| Dormant (>365d) | N |\n\n## Code source status\n`forks_list=ok|fail · compare_ok=N\u002FM · deep_read=N\u002FM · rate_limit_retries=N · unreadable=N`\n```\n\nIf PROMOTE has >5 forks, keep only the top 5 by score; list the rest in REVIEW.\n\n### A8. Update code state\n\nWrite `memory\u002Ftopics\u002Ffork-fleet-state.json`:\n\n```json\n{\n  \"last_run\": \"${today}\",\n  \"last_status\": \"FORK_FLEET_OK\",\n  \"parent_repo\": \"owner\u002Frepo\",\n  \"forks\": {\n    \"owner\u002Frepo\": {\n      \"pushed_at\": \"YYYY-MM-DD...\",\n      \"default_branch\": \"main\",\n      \"ahead_by\": N,\n      \"behind_by\": N,\n      \"new_skill_count\": N,\n      \"score\": N,\n      \"tier\": \"PROMOTE|REVIEW|NOTE|UNREADABLE|API_FAIL\",\n      \"unique_skills\": [\"skills\u002Ffoo\u002FSKILL.md\", \"...\"]\n    }\n  }\n}\n```\n\n### A9. Set the code branch status\n\n| Status | Meaning |\n|--------|---------|\n| `FORK_FLEET_OK` | Active forks present AND (PROMOTE\u002FREVIEW non-empty OR delta non-empty) → contributes a notify signal |\n| `FORK_FLEET_NO_CHANGE` | Active forks exist but nothing crossed REVIEW and delta is empty → log only |\n| `FORK_FLEET_QUIET` | Zero active forks and no state change → log only |\n| `FORK_FLEET_API_FAIL` | Fork listing failed or >50% of compares failed → error signal |\n\n---\n\n## Branch B — Config divergence (runs when `SCOPE ∈ {config, both}`)\n\n### B1. Snapshot upstream defaults\n\nRead this running instance's local `aeon.yml` once. Build (these are baselines — never mutated):\n\n- `UPSTREAM_DEFAULTS`: dict `{skill_name -> {enabled: bool, model: str|null, var: str, schedule: str|null}}` for every skill entry under `skills:`.\n- `UPSTREAM_SKILLS`: set of skill directory names from `skills\u002F` (use `ls skills\u002F`).\n- `UPSTREAM_TAGS`: dict `{skill_name -> [tags]}` parsed from each `skills\u002F\u003Cname>\u002FSKILL.md` frontmatter (best-effort; missing frontmatter → `[]`).\n\n### B2. Per-fork enumeration (one tree call + one yml fetch each)\n\nOperate over the **active-fork set** from shared step S3 (already filtered to forks pushed in the last 30 days - the config branch's original cutoff). For each active fork, run **one** recursive git-tree call to enumerate files (cheaper than per-path contents), one call per fork with the fork written in literally:\n\n```bash\ngh api \"repos\u002F\u003CFORK_FULL>\u002Fgit\u002Ftrees\u002FHEAD?recursive=1\" --jq '[.tree[] | select(.type == \"blob\") | .path]'\n```\n\nThen fetch the fork's `aeon.yml` only if the tree contains it:\n\n```bash\ngh api \"repos\u002F\u003CFORK_FULL>\u002Fcontents\u002Faeon.yml?ref=\u003CFORK_DEFAULT_BRANCH>\" --jq '.content' | base64 -d\n```\n\nError handling:\n- 404 \u002F 409 (empty repo): mark `status: \"no_tree\"`, skip aeon.yml extraction, continue.\n- 403 with `X-RateLimit-Remaining: 0`: sleep 60s, retry once. If still failing, mark `status: \"rate_limited\"` and continue.\n- Tree contains aeon.yml but the contents call 404s: mark `status: \"yml_unreadable\"`, continue.\n- aeon.yml present but YAML parse fails: mark `status: \"yml_invalid\"`, continue.\n\nFor each readable `aeon.yml`, extract per-skill `{enabled, model, var, schedule}`. Treat missing keys as inheriting the upstream default (do **not** count those as overrides).\n\nDetect **fork-only skills**: directory names matching `skills\u002F\u003Cname>\u002FSKILL.md` in the fork's tree where `\u003Cname>` is NOT in `UPSTREAM_SKILLS`. Record `{fork_full_name, skill_name, path}` for each.\n\n### B3. Tier each fork\n\nCompute a divergence signal vector vs `UPSTREAM_DEFAULTS`:\n- `enabled_diff`: count of skills where the fork's `enabled` differs from upstream\n- `var_overrides`: count of skills with non-empty `var:` where upstream's was empty (or a different non-empty value)\n- `model_overrides`: count of skills with `model:` differing from upstream\n- `schedule_overrides`: count of skills with `schedule:` differing from upstream\n- `fork_only_skill_count`: count from step B2\n\nTier the fork:\n- **CONFIGURED**: any signal ≥1 (the fork actively diverged)\n- **TEMPLATE**: aeon.yml readable but every signal is 0 — excluded from divergence math\n- **UNREADABLE**: no_tree \u002F no aeon.yml \u002F yml_unreadable \u002F yml_invalid \u002F rate_limited — tracked in the source-status footer\n\nLet `N_CONFIGURED` = count of forks tiered CONFIGURED. **If `N_CONFIGURED \u003C 2`:** the config branch cannot produce meaningful divergence math. Set config status `FORK_SKILL_DIGEST_TEMPLATE_FLEET`, record active\u002Ftemplate\u002Funreadable counts, emit a stub Part 2 noting the conversion rate, and contribute **no** config notify signal. Skip steps B4–B6.\n\n### B4. Aggregate divergence (the core config analysis)\n\nFor each skill name in `UPSTREAM_SKILLS`, compute four dimensions:\n\n**Enable divergence:**\n- `forks_enabled_count`: number of CONFIGURED forks with `enabled: true` for this skill\n- `forks_disabled_count`: number of CONFIGURED forks with `enabled: false` (explicitly set, not inherited)\n- `upstream_enabled`: bool from UPSTREAM_DEFAULTS\n- `divergence_pct`:\n  - If upstream `enabled: false`: `forks_enabled_count \u002F N_CONFIGURED` (how many disagree by enabling)\n  - If upstream `enabled: true`: `forks_disabled_count \u002F N_CONFIGURED` (how many disagree by disabling)\n- `direction`: `\"ENABLE_UPWARD\"` (upstream off, forks turn on) or `\"DISABLE_DOWNWARD\"` (upstream on, forks turn off)\n\n**Var divergence:**\n- `var_override_count`: number of CONFIGURED forks where `var:` differs from upstream\n- `top_var_value`: most common non-empty fork value (with count) — only if ≥2 forks share it\n\n**Model divergence:**\n- `model_override_count`: number of forks with non-null model differing from upstream\n- `top_model_value`: most common fork model (with count) — only if ≥2 forks share it (signals fleet consensus on a cheaper\u002Fdifferent model)\n\n**Schedule divergence:**\n- `schedule_override_count`: number of forks with schedule differing from upstream\n- `top_schedule_value`: most common fork schedule (with count) — only if ≥2 forks share it\n\n### B5. Categorize divergent skills\n\nClassify each skill into **at most one** bucket (first match wins, in this order):\n\n- **DEFAULT_FLIP_ENABLE**: `direction == \"ENABLE_UPWARD\"` AND `divergence_pct >= 0.50` AND skill is not `workflow_dispatch` AND skill not tagged `meta`\u002F`dev`. Recommend: flip upstream default to `enabled: true`.\n- **DEFAULT_FLIP_DISABLE**: `direction == \"DISABLE_DOWNWARD\"` AND `divergence_pct >= 0.50`. Recommend: flip upstream default to `enabled: false` (the fleet is voting it as noise).\n- **MODEL_CONSENSUS**: `top_model_value` non-null AND its count `>= max(2, ceil(N_CONFIGURED * 0.40))`. Recommend: match the fleet's model in upstream.\n- **VAR_HOTSPOT**: `var_override_count >= max(2, ceil(N_CONFIGURED * 0.30))` AND `top_var_value` non-null. Recommend: surface the common var value in upstream docs or as the default.\n- **EMERGING**: `direction == \"ENABLE_UPWARD\"` AND `0.25 \u003C= divergence_pct \u003C 0.50` AND not already in a flip bucket. Surface as a watchlist — fleet sentiment building but not yet majority.\n- (otherwise: not categorized; appears only in the appendix divergence table if any signal is non-zero)\n\nSkills with all-zero divergence are omitted.\n\n### B6. Per-fork customization fingerprint\n\nFor each CONFIGURED fork:\n- `total_overrides`: `enabled_diff + var_overrides + model_overrides + schedule_overrides + fork_only_skill_count`\n- `category_lean`: dict `{tag -> count_of_enabled_skills_with_that_tag}` (using UPSTREAM_TAGS for upstream skills the fork enables; fork-only skills counted under tag `\"fork-only\"`)\n- `dominant_category`: tag with max count, or `\"mixed\"` if no tag holds >40% of total enabled count\n\nRank forks by `total_overrides` desc. Top 5 = \"heaviest customizers\" - surface with dominant category and a one-line synthesis (e.g. `\"owner\u002Faeon - content-heavy: 14 article\u002Fdigest skills enabled, 3 model overrides to claude-sonnet-5-5\"`). The fingerprint is **descriptive only** - never recommend changes to individual forks.\n\n### B7. Config week-over-week delta\n\nRead the prior `memory\u002Ftopics\u002Ffork-digest-state.json` snapshot (schema in B8). If it exists and `last_run` is within the last 14 days, compute:\n- **NEW_FLIP**: skills now in DEFAULT_FLIP_* that weren't last run\n- **STRENGTHENED**: skills that moved EMERGING → DEFAULT_FLIP_ENABLE\n- **FADED**: skills that left a flip bucket since last run\n- **NEW_FORK_ONLY**: fork-only skills not present last run\n- **NEW_HEAVY_CUSTOMIZER**: forks now in the top-5 fingerprint that weren't before\n\nIf the file is missing or stale (>14 days), set all deltas to `\"first divergence snapshot\"`.\n\n### B8. Pick the config verdict + persist snapshot\n\nConfig verdict line, strongest single claim first:\n1. Any `DEFAULT_FLIP_ENABLE`: `\"${N} forks enable ${skill} (upstream defaults off) — flip the default\"`\n2. Else any `DEFAULT_FLIP_DISABLE`: `\"${N} forks disable ${skill} (upstream defaults on) — fleet is voting it as noise\"`\n3. Else any `MODEL_CONSENSUS`: `\"${N} forks override ${skill} → ${model} — match upstream\"`\n4. Else any `NEW_FORK_ONLY` from delta: `\"${fork_owner} shipped ${skill} — not in upstream\"`\n5. Else any `EMERGING`: `\"${skill} adoption building (${pct}% of configured) — watchlist\"`\n6. Else: `\"${N_CONFIGURED} configured forks; no divergence pattern crossed flip threshold\"`\n\nPersist `memory\u002Ftopics\u002Ffork-digest-state.json` (overwrite each run — the JSON is the delta contract; do NOT parse last week's article):\n\n```json\n{\n  \"last_run\": \"${today}\",\n  \"target_repo\": \"${PARENT_REPO}\",\n  \"n_active\": N_ACTIVE,\n  \"n_configured\": N_CONFIGURED,\n  \"n_template\": N_TEMPLATE,\n  \"n_unreadable\": N_UNREADABLE,\n  \"buckets\": {\n    \"DEFAULT_FLIP_ENABLE\": [{\"skill\": \"name\", \"forks\": N, \"pct\": 0.NN}],\n    \"DEFAULT_FLIP_DISABLE\": [{\"skill\": \"name\", \"forks\": N, \"pct\": 0.NN}],\n    \"MODEL_CONSENSUS\": [{\"skill\": \"name\", \"model\": \"value\", \"forks\": N}],\n    \"VAR_HOTSPOT\": [{\"skill\": \"name\", \"var\": \"value\", \"forks\": N}],\n    \"EMERGING\": [{\"skill\": \"name\", \"pct\": 0.NN}]\n  },\n  \"fork_only_skills\": [{\"fork\": \"owner\u002Frepo\", \"skill\": \"name\"}],\n  \"fingerprints\": [{\"fork\": \"owner\u002Frepo\", \"total_overrides\": N, \"dominant_category\": \"tag\"}]\n}\n```\n\n### B9. Build the config-divergence article part\n\nAssemble this block (it becomes **Part 2** of the combined article):\n\n```markdown\n*Scanned ${N_ACTIVE} active forks of ${PARENT_REPO} (pushed in last 30 days). ${N_CONFIGURED} are configured (aeon.yml diverges from upstream defaults). Divergence scored against the configured ${N_CONFIGURED}.*\n\n## Default-flip candidates\n\n### Enable upward (upstream off → fleet enables)\n| Skill | Forks enabled | % of configured | Δ vs last week |\n|-------|---------------|-----------------|----------------|\n| name  | N             | XX%             | NEW \u002F STRENGTHENED \u002F - |\n\n(Only DEFAULT_FLIP_ENABLE. If empty: \"No skills crossed the 50% enable-upward threshold this week.\")\n\n### Disable downward (upstream on → fleet disables)\n| Skill | Forks disabled | % of configured | Δ vs last week |\n|-------|----------------|-----------------|----------------|\n| name  | N              | XX%             | NEW \u002F - |\n\n(Only DEFAULT_FLIP_DISABLE. If empty: \"No skills crossed the 50% disable-downward threshold.\")\n\n## Fleet consensus on alternative settings\n\n### Model overrides\n${MODEL_CONSENSUS entries: \"skill X - N forks → claude-sonnet-5-5 (40% of configured)\" OR \"none this week\"}\n\n### Var hotspots\n${VAR_HOTSPOT entries: \"skill X - N forks set var to '${value}'\" OR \"none this week\"}\n\n### Schedule overrides\n${skills where ≥2 forks share an alternative schedule, with the schedule string OR \"none this week\"}\n\n## Watchlist (emerging - 25-49% adoption)\n${EMERGING skills with adoption % OR \"none this week\"}\n\n## Heaviest customizers (top 5)\n\n| Fork | Total overrides | Dominant category | Notes |\n|------|-----------------|-------------------|-------|\n| owner\u002Frepo | N | content \u002F dev \u002F meta \u002F fork-only \u002F mixed | one-line synthesis |\n\n## Fork-only skills\n\n${list of {fork, skill_name} pairs OR \"none this week\"}\n\n(These skills exist as `skills\u002F\u003Cname>\u002FSKILL.md` in a fork but not in upstream - fork experiments worth reviewing for upstreaming.)\n\n## Config week-over-week\n\n${\"First divergence snapshot - no comparison\" OR list of NEW_FLIP \u002F STRENGTHENED \u002F FADED \u002F NEW_FORK_ONLY \u002F NEW_HEAVY_CUSTOMIZER}\n\n## Fleet composition (config tiers)\n\n| Tier | Count | % |\n|------|-------|---|\n| Configured | N_CONFIGURED | XX% |\n| Template (untouched aeon.yml) | N_TEMPLATE | XX% |\n| Unreadable | N_UNREADABLE | XX% |\n| **Total active** | N_ACTIVE | 100% |\n\n## Config source status\n- Trees fetched: N_TREES_OK \u002F N_ACTIVE\n- aeon.yml readable: (N_CONFIGURED + N_TEMPLATE) \u002F N_ACTIVE\n- YAML parse failures: N_YML_INVALID\n- Rate-limited: N_RATE_LIMITED\n- Fork-only skills inspected: N_FORK_ONLY_FILES\n\n## Appendix - full divergence table\n\n(Every skill with ≥1 non-zero divergence signal, sorted by total override count desc. Columns: skill, enable_diff, var_overrides, model_overrides, schedule_overrides. Cap at 30 rows; if more, append \"+ N more skills with low-signal divergence\".)\n```\n\n### Config branch status\n\n| Status | Meaning |\n|--------|---------|\n| `FORK_SKILL_DIGEST_OK` | ≥2 configured forks AND ≥1 flip\u002Fconsensus\u002Fnew-fork-only signal → contributes a notify signal |\n| `FORK_SKILL_DIGEST_QUIET` | ≥2 configured forks but no signal crossed thresholds → log only |\n| `FORK_SKILL_DIGEST_TEMPLATE_FLEET` | \u003C2 configured forks (mostly templates) → log only |\n| `FORK_SKILL_DIGEST_NO_FORKS` | Zero active forks → log only |\n\n---\n\n## Assemble the report (all scopes)\n\n### R1. Write the combined article\n\nTo `output\u002Farticles\u002Ffork-divergence-${today}.md`. Header first, then whichever parts ran:\n\n```markdown\n# Fork Divergence - ${today}\n\n**Verdict:** {lead with the stronger of the two sub-verdicts - a code PROMOTE\u002FNEW-UPSTREAM-CANDIDATE outranks a config flip only if it's a genuinely new skill; otherwise a DEFAULT_FLIP leads. Use judgment; one line.}\n\n- **Code divergence:** {code verdict from A6, or \"not run (scope=config)\"}\n- **Config divergence:** {config verdict from B8, or \"not run (scope=code)\"}\n\nFleet: N_TOTAL total forks · N_ACTIVE active · N_MANAGED managed instances · N_COMMUNITY community.\n\n---\n\n# Part 1 - Code divergence\n{Branch A article part from A7; omit this whole part if SCOPE=config}\n\n---\n\n# Part 2 - Config divergence\n{Branch B article part from B9; omit this whole part if SCOPE=code}\n\n---\n*Source: GitHub API - forks of ${PARENT_REPO}. Code divergence = per-fork unique commits\u002Fskills vs upstream. Config divergence = where configured forks' aeon.yml disagrees with upstream `enabled`\u002F`var`\u002F`model`\u002F`schedule`; untouched templates are excluded from the config math. Companion to `skill-gap` (popularity).*\n```\n\nCap the article at ~700 lines total (≈500 for code sections when both parts run). When only one branch ran, drop the other Part heading and the missing sub-verdict line.\n\n### R2. Notify — gated\n\nRead `soul\u002F` (if present) to match the operator's voice. **Skip notify entirely** when neither branch produced a signal, i.e. when:\n- code status ∈ {`FORK_FLEET_NO_CHANGE`, `FORK_FLEET_QUIET`} (or code didn't run), **AND**\n- config status ∈ {`FORK_SKILL_DIGEST_QUIET`, `FORK_SKILL_DIGEST_TEMPLATE_FLEET`, `FORK_SKILL_DIGEST_NO_FORKS`} (or config didn't run).\n\nIf either branch hit `FORK_FLEET_API_FAIL`, send an **error** notify (`--severity warn`) noting the failure and source status.\n\nOtherwise send one combined message via `.\u002Fnotify` (include only the sub-blocks whose branch produced signal; keep it tight):\n\n```\n*Fork Divergence - ${today}*\n{combined verdict line}\n\nFleet: N_ACTIVE active \u002F N_TOTAL total. {1 sentence on shape - \"mostly managed instances\", \"community picking up\", \"template-heavy\", etc.}\n\n{If code PROMOTE non-empty:}\nUpstream candidate: {top PROMOTE fork}\n{2 sentences: what they built, why it's worth merging back}\n\n{If code delta has NEW_SKILLS:}\nNew skills landed this week:\n- {fork} → `skills\u002Ffoo\u002FSKILL.md` - {synthesis}\n\n{If DEFAULT_FLIP_ENABLE non-empty (top 3):}\nFlip enable (upstream off → fleet on):\n- {skill} - {N} forks ({pct}%)\n\n{If DEFAULT_FLIP_DISABLE non-empty (top 3):}\nFlip disable (upstream on → fleet off):\n- {skill} - {N} forks ({pct}%)\n\n{If MODEL_CONSENSUS non-empty (top 2):}\nModel consensus:\n- {skill} → {model} ({N} forks)\n\n{If config delta NEW_FORK_ONLY non-empty:}\nNew fork-only skills: {comma-separated owner\u002Fskill, capped at 3}\n\nFull report: https:\u002F\u002Fgithub.com\u002F\u003CSELF_REPO>\u002Fblob\u002Fmain\u002Foutput\u002Farticles\u002Ffork-divergence-${today}.md\n```\n\nUse this running instance's repo (`GITHUB_REPOSITORY`) for the URL (the article lives in this running instance's repo, not the target repo). It is SELF_REPO, printed by `gh repo view --json nameWithOwner -q .nameWithOwner` (run that as its own call if S2 did not already); write it in literally.\n\n### R3. Log\n\nAppend to `memory\u002Flogs\u002F${today}.md` under **one** heading. Include a discriminator line naming the scope that ran, then only the sub-blocks for branches that ran:\n\n```\n### fork-fleet\n- Scope: {both | code | config}  ·  Combined status: {FORK_DIVERGENCE_OK | NO_CHANGE | QUIET | NO_TARGET | API_FAIL}\n- Verdict: {combined verdict line}\n- Fleet: N_ACTIVE active \u002F N_TOTAL total (N_MANAGED managed, N_COMMUNITY community)\n\n[code]  (only if Branch A ran)\n- Code status: {FORK_FLEET_OK | NO_CHANGE | QUIET | API_FAIL}\n- PROMOTE: N forks (list), REVIEW: N, NOTE: N\n- Code delta: {new_forks:N, new_active:N, new_skills:N, went_stale:N}\n- Code source: forks_list=ok|fail · compare_ok=N\u002FM · deep_read=N\u002FM · unreadable=N\n\n[config]  (only if Branch B ran)\n- Config status: {FORK_SKILL_DIGEST_OK | QUIET | TEMPLATE_FLEET | NO_FORKS}\n- Configured: N (XX% conversion) · Template: N · Unreadable: N\n- DEFAULT_FLIP_ENABLE: N · DEFAULT_FLIP_DISABLE: N · MODEL_CONSENSUS: N · VAR_HOTSPOT: N · EMERGING: N\n- Fork-only skills: N · Heaviest customizer: {fork} ({N} overrides)\n\n- Article: output\u002Farticles\u002Ffork-divergence-${today}.md\n- Notification sent: yes\u002Fno\n```\n\n## Exit taxonomy (combined)\n\nThe combined status rolls up the per-branch statuses (kept verbatim in A9 \u002F B-status above):\n\n| Combined status | Rolls up when | Notify? |\n|-----------------|---------------|---------|\n| `FORK_DIVERGENCE_OK` | code = `FORK_FLEET_OK` **OR** config = `FORK_SKILL_DIGEST_OK` | Yes |\n| `FORK_DIVERGENCE_NO_CHANGE` | both branches ran but neither reached OK (all `NO_CHANGE`\u002F`QUIET`\u002F`TEMPLATE_FLEET`\u002F`NO_FORKS`) | No (log only) |\n| `FORK_DIVERGENCE_QUIET` | zero active forks and no code-side state change (S3 short-circuit) | No (log only) |\n| `FORK_DIVERGENCE_NO_TARGET` | no parent\u002Ftarget repo resolved (S2) | No (log only) |\n| `FORK_DIVERGENCE_API_FAIL` | fork listing failed, or >50% of a branch's compares\u002Ftrees failed | Yes (error notify) |\n\n## Constraints\n\n- **Cross-repo compare** accepts up to 300 files per response; if a fork exceeds this, note `files_truncated=true` for it and proceed.\n- Cap active-fork deep processing at **50 per run** (S3) — rank by `pushed_at_epoch` desc and trim, logging `truncated_at=50`.\n- Never deep-read content from a fork with `archived=true`, or when the SKILL.md path is absent from the compare `files` list (cheapest sanity check).\n- **Never invent a PROMOTE candidate** — a fork with zero new skill files is at most REVIEW.\n- **Config math needs a denominator:** never send a config signal when `N_CONFIGURED \u003C 2` — the divergence percentages are meaningless without a configured base.\n- Skills tagged `meta` or `dev` are excluded from `DEFAULT_FLIP_ENABLE` (operator tools — fork adoption isn't the success metric). They can still appear in MODEL_CONSENSUS, VAR_HOTSPOT, and the appendix.\n- Skills with `schedule: \"workflow_dispatch\"` are excluded from **both** flip buckets (on-demand by design — adoption % is misleading).\n- `heartbeat` is excluded from `DEFAULT_FLIP_DISABLE` (every fork inheriting upstream's `enabled: true` would game the disable count if any fork explicitly disables it to stay quiet).\n- The per-fork fingerprint is descriptive only — only aggregate signals drive recommendations.\n- Silent runs are **correct**, not failures. This skill is the divergence companion to `skill-gap` (popularity); avoid duplicating its headline metrics — focus on the code + config **divergence patterns** it doesn't surface.\n\n## Network note\n\nEvery GitHub call uses `gh api`, which authenticates via `GITHUB_TOKEN` automatically — no `curl`, no `$SECRET` on the command line (so nothing for the Bash permission layer to refuse), no secrets beyond the default `GITHUB_TOKEN`. Retry policy: on `429`\u002F`5xx` (compare) back off per step A1; on `403` with `X-RateLimit-Remaining: 0` (tree\u002Fcontents) sleep 60s and retry once, then mark that fork `rate_limited` and proceed with a partial fleet (the verdict and source-status footers surface the gap). If the initial `\u002Fforks` listing fails after retry, combined status = `FORK_DIVERGENCE_API_FAIL` with `forks_list=fail`.\n",{"name":5,"metadata":43,"description":7},{"var":44,"tags":45,"category":35},"",[33,46],"meta","main","504a21d53c0c9b485f6e90c43db83000825e1d3c","skills\u002Ffork-fleet",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":57,"prompt":67,"zipAvailable":68},[58,61,64],{"tool":59,"command":60},"npx","npx skills add aeonfun\u002Faeon --skill fork-fleet",{"tool":62,"command":63},"bunx","bunx skills add aeonfun\u002Faeon --skill fork-fleet",{"tool":65,"command":66},"pnpm","pnpm dlx skills add aeonfun\u002Faeon --skill fork-fleet","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Faeonfun\u002Faeon\u002Fblob\u002Fmain\u002Fskills\u002Ffork-fleet\u002FSKILL.md",true,[],[71,78,84,90,96,102],{"skillNo":72,"name":73,"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":74,"categories":75,"keywords":76,"publishedAt":77},"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":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":77},"01M7DSM0496DRYYW5A4NAAHZDX","Warn Hidden CSS","https:\u002F\u002Fgithub.com\u002Faeonfun\u002Faeon\u002Fblob\u002Fmain\u002Fscripts\u002Ftests\u002Fskill-scan-fixtures\u002Fwarn-hidden-css\u002FSKILL.md",[],[],{"skillNo":85,"name":86,"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":87,"categories":88,"keywords":89,"publishedAt":77},"01M7DSKWCGDVYZSDJM31YHS2JB","Malicious Injection Uppercase","https:\u002F\u002Fgithub.com\u002Faeonfun\u002Faeon\u002Fblob\u002Fmain\u002Fscripts\u002Ftests\u002Fskill-scan-fixtures\u002Fmalicious-injection-uppercase\u002FSKILL.md",[],[],{"skillNo":91,"name":92,"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":93,"categories":94,"keywords":95,"publishedAt":77},"01M7DSKRV2RRA1V0HK13KYWJXD","Malicious Injection Homoglyph","https:\u002F\u002Fgithub.com\u002Faeonfun\u002Faeon\u002Fblob\u002Fmain\u002Fscripts\u002Ftests\u002Fskill-scan-fixtures\u002Fmalicious-injection-homoglyph\u002FSKILL.md",[],[],{"skillNo":97,"name":98,"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":99,"categories":100,"keywords":101,"publishedAt":77},"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":103,"name":104,"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":105,"categories":106,"keywords":107,"publishedAt":77},"01M7DSKGFQ5V2VNSA4K97CVFMB","Benign Greek Math","https:\u002F\u002Fgithub.com\u002Faeonfun\u002Faeon\u002Fblob\u002Fmain\u002Fscripts\u002Ftests\u002Fskill-scan-fixtures\u002Fbenign-greek-math\u002FSKILL.md",[],[]]