[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M71RFG1VEMT3F1ZZ1E712ZH8":3,"skill-same:01M71RFG1VEMT3F1ZZ1E712ZH8":114},{"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":23,"publishedAt":44,"linkOnly":45,"markdownContent":46,"markdownBody":47,"frontMatter":48,"refName":49,"commitSha":50,"skillDirectory":51,"forksCount":52,"repositoryHtmlUrl":53,"repositoryHomepageUrl":6,"ownerHtmlUrl":54,"ownerBlogUrl":6,"ownerDescription":55,"updatedAt":56,"install":57,"resources":70},"01M71RFG1VEMT3F1ZZ1E712ZH8","autospec-listen",null,"Use when the user expresses an imperative intent to file an issue, write a spec, design a feature, implement\u002Fbuild\u002Fship something, review code, or run autospec mid-conversation. Trigger keywords are \"file an issue\", \"new issue\", \"open an issue\", \"create a ticket\", \"write a spec\", \"design spec\", \"new spec\", \"start a spec\", \"design\", \"new feature\", \"implement\", \"build\", \"ship\", \"review\", \"autospec\". On an \"issue\" trigger, drafts a body from the last ~10 conversation turns and asks the user to approve before running gh issue create. On a \"spec\" trigger, hands off to \u002Fautospec-define. On a build\u002Fchange verb, gates intent via the deterministic listener-match classifier and auto-routes to the mapped autospec skill with a one-line opt-out. Lives at github.com\u002Fberlinguyinca\u002Fautospec\u002Fskills\u002Fautospec-listen.","berlinguyinca\u002Fautospec","berlinguyinca","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F16364?v=4","01M71R7C0SWBG05FNHQE4XQ7Y3","Gert Wohlgemuth","01M71R7C1VY5K3Q715RNDDBVGQ",3,"MIT","https:\u002F\u002Fgithub.com\u002Fberlinguyinca\u002Fautospec\u002Fblob\u002Fmain\u002Fskills\u002Fautospec-listen\u002FSKILL.md",[18],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000023","Skill与智能体开发","SCENARIO","使用场景",[24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43],"autospec","listen","autospec listen","intent","change","write a spec","create","design spec","review","code","when","auto","issue create","new","design","open","ticket","listener","verb","conversation","2026-11-10 13:02:02",false,"---\nname: autospec-listen\ndescription: |\n  Use when the user expresses an imperative intent to file an issue, write a spec, design a feature, implement\u002Fbuild\u002Fship something, review code, or run autospec mid-conversation. Trigger keywords are \"file an issue\", \"new issue\", \"open an issue\", \"create a ticket\", \"write a spec\", \"design spec\", \"new spec\", \"start a spec\", \"design\", \"new feature\", \"implement\", \"build\", \"ship\", \"review\", \"autospec\". On an \"issue\" trigger, drafts a body from the last ~10 conversation turns and asks the user to approve before running gh issue create. On a \"spec\" trigger, hands off to \u002Fautospec-define. On a build\u002Fchange verb, gates intent via the deterministic listener-match classifier and auto-routes to the mapped autospec skill with a one-line opt-out. Lives at github.com\u002Fberlinguyinca\u002Fautospec\u002Fskills\u002Fautospec-listen.\n---\n\n# autospec-listen (harness-neutral)\n\nConversation listener that watches for issue-filing or spec-writing intent\nmid-conversation and dispatches to the right downstream flow without forcing\nthe user to remember a slash command.\n\n\u003C!-- autospec-block:startup-self-update SKILL_NAME=autospec-listen -->\n\n## Self-update mode\n\nIf the feature-request argument matches the regex `^\\s*update\\s*$` (case-insensitive, whitespace-padded), this skill enters self-update mode and does not run the normal trigger flow:\n\n1. **Detect harness** by checking which install path exists for this skill:\n   - Claude Code: `~\u002F.claude\u002Fskills\u002Fautospec-listen\u002FSKILL.md`\n   - OpenCode:    `~\u002F.config\u002Fopencode\u002Fagent\u002Fautospec-listen.md`\n   - Codex CLI:   `~\u002F.codex\u002Fprompts\u002Fautospec-listen.md`\n2. **Re-install the full autospec suite from `main`** by piping the canonical installer:\n   ```bash\n   curl -fsSL https:\u002F\u002Fraw.githubusercontent.com\u002Fberlinguyinca\u002Fautospec\u002Fmain\u002Fbootstrap.sh | bash -s -- --skill all --harness all --update\n   ```\n   Run this one-liner once; it refreshes all autospec skills across all harnesses.\n3. **Show the diff** between the prior installed file(s) and the freshly fetched copy (e.g. `diff \u003C(cat \u003Cprior>) \u003C(curl -fsSL ...SKILL.md)` or the equivalent recorded by the installer).\n4. **Stop.** Do not enter any trigger flow. Print the upgrade summary and return to the user.\n\nIf no install path is detected, print `Self-update: no installed copy of autospec-listen found; run install.sh first.` and exit.\n\n## Harness detection (run once at skill start)\n\nDetect your harness by checking available tools before dispatching work:\n\n1. **Claude Code** — the `Agent` tool with a `subagent_type` parameter is available.\n   - `TIER_A` = `opus` + `ultrathink`  (the alias, which resolves to the current Claude Opus generation)\n   - `TIER_B` = `sonnet`               (the alias, which resolves to the current Claude Sonnet generation)\n\n2. **OpenCode** — a `task` tool with model\u002Ftier configuration is available (no `subagent_type`).\n   - `TIER_A` = top-tier task model + high reasoning\n   - `TIER_B` = smaller-tier task model + medium reasoning\n\n3. **Codex CLI** — neither `Agent` nor a configurable `task` tool is available; `apply_patch` is the primary edit tool.\n   - `TIER_A` = current top GPT model + `reasoning_effort=high`\n   - `TIER_B` = the spark \u002F cost-optimized variant of the configured model when one exists, else the configured model + `reasoning_effort=medium`\n\n**Fallback rule:** If `TIER_B` is not available in your harness (model unknown, quota\u002Fcapacity failure, authorization failure, or tool call returns an error for that model), silently retry the same dispatch with `TIER_A`. Preserve the parent context on retry; for Codex native subagents, fork\u002Finherit the current conversation context and use the latest top GPT model instead of moving the work into the main session. Never ask the user.\n\nHold `TIER_A` and `TIER_B` for the entire skill run.\n\n## Required capabilities & harness adapter\n\nThis skill assumes four capabilities. Map each one to your harness's actual tool. If a capability is missing, use the listed fallback.\n\n| Capability                  | Claude Code                          | OpenCode                                 | Codex CLI                                | Fallback if missing                                |\n|-----------------------------|--------------------------------------|------------------------------------------|------------------------------------------|----------------------------------------------------|\n| Read recent conversation    | harness-native message buffer        | harness-native message buffer            | harness-native message buffer            | Ask the user to paste the relevant turns           |\n| Ask the user a question     | `AskUserQuestion`                    | inline prompt                            | inline prompt                            | Ask in the response and wait for the next turn     |\n| Run `gh` CLI                | shell tool                           | shell tool                               | shell tool                               | Have the user run the command and confirm          |\n| Hand off to another skill   | `Skill` invocation (`\u002Fautospec-define`) | nested skill invocation                | `\u002Fautospec-define` slash command         | Print the suggested command for the user to run    |\n| Subagent model tier         | Tier B: `sonnet` + medium thinking; Tier A only if escalating to \u002Fautospec-define which honors its own tier policy | Tier B: smaller-tier `task` + medium reasoning; escalation honors target skill | Tier B: current spark\u002Fcost-optimized Codex + `reasoning_effort=medium`; escalation honors target | Honor AGENTS.md tier mapping; retry the same subagent UP on unavailability |\n\u003C!-- autospec-block:harness-adapter-core -->\n\n**Persistent project notes**: write durable preferences to **`AGENTS.md`** in the repo root — this is the de-facto standard recognized by Claude Code (also reads `CLAUDE.md`), OpenCode, and Codex. Per AGENTS.md, this listener runs at **Tier B** (it's a router, not a designer); when it hands off to `\u002Fautospec-define` for a spec trigger, the downstream skill applies its own tier policy. Fall back UP the tier on quota\u002Fcapacity or other unavailability by retrying the same subagent with the stronger tier while preserving parent context.\n\n## Trigger flow\n\n> **Model tier:** Tier B (implementation work) — this listener is a router; subagent dispatches inside its trigger flows (drafting issue bodies, summarizing recent conversation) use cheaper models per AGENTS.md. Spec-trigger handoff to `\u002Fautospec-define` lets the downstream skill apply its own tier policy.\n\n## Keyword auto-routing\n\nBeyond the explicit issue\u002Fspec triggers, this listener routes common build\u002Fchange verbs into the matching autospec skill so the pipeline is the default path. Routing is gated by an imperative-intent check and always offers a one-line opt-out.\n\n**Intent gate.** Do not classify by eye. Delegate to the deterministic classifier, passing the user's latest message verbatim:\n\n```bash\nbash \"${AUTOSPEC_SCRIPTS_DIR:-$HOME\u002F.autospec\u002Fscripts}\u002Flistener-match.sh\" --classify \"\u003Cuser message>\"\n```\n\nIt emits `{\"match\":bool,\"skill\":...,\"trigger\":...,\"intent\":\"imperative|incidental|none\",\"confidence\":0-1,\"autonomous\":bool,\"chain\":\u003Cskill|null>,\"gate\":\u003Cstring|null>}`. The gate is biased to false-negatives: descriptive (\"the design is nice\"), past-tense (\"I reviewed it\"), negated (\"don't implement yet\"), and interrogative (\"should we redesign?\") uses return `match:false` and MUST NOT route. Prefer no route over a misfire.\n\nThe two new trailing fields:\n- `chain` — after the primary skill completes, invoke the chained skill (e.g. `\u002Fautospec-run` after `\u002Fautospec-refine`). Null when no chain applies.\n- `gate` — before routing, the listener must satisfy the named context gate; `auto-implement-open` means run `gh issue list --repo \u003Crepo> --label auto-implement --state open --json number --jq 'length'` and route only if > 0, else stay in plain mode. `explore-confirm` means the route launches `\u002Fautospec-explore`, which starts a perpetual autonomous research + ship loop on an isolated sandbox branch whose PRs target the sandbox branch, never main; before invoking, print exactly that consequence and require ONE explicit confirmation, routing only on an affirmative reply. Null means no gate.\n\n**Verb → skill map (D3).**\n\n| Verb(s) in an imperative request | Routes to |\n|----------------------------------|-----------|\n| `design` \u002F `new feature` \u002F `spec` | `\u002Fautospec-define` |\n| `implement` \u002F `build` \u002F `ship` | `\u002Fautospec-run` (or `\u002Fautospec` end-to-end when no issues exist yet) |\n| `refine` \u002F `optimize` \u002F `polish` \u002F `improve` \u002F `tune` | `\u002Fautospec-refine` |\n| `run` (scoped: \"run it\", \"run the loop\", \"run autospec\"; bare \"run\" only when open `auto-implement` issues exist) | `\u002Fautospec-run` |\n| `refine and run` \u002F `tune up` (combined) | `\u002Fautospec-refine` then `\u002Fautospec-run` |\n| `review` | `\u002Fautospec-review` |\n| `explore` \u002F `discover` (only with an explicit build\u002Fship\u002Fimprove action connector, e.g. \"explore and ship\", \"discover features to build\") | `\u002Fautospec-explore` (gate `explore-confirm`) |\n| `fix` (imperative, e.g. \"fix the login flow\"; trivial-fix phrasings like \"fix the typo\", \"quick fix\", \"fix the formatting\" stay plain) | the umbrella `\u002Fautospec` (no gate) |\n| `autospec …` | the umbrella `\u002Fautospec` |\n| GitHub Projects board URL (`https:\u002F\u002Fgithub.com\u002F(orgs\\|users)\u002F\u003Cname>\u002Fprojects\u002F\u003Cn>`, optional `\u002Fviews\u002F\u003Cn>`) **with** a build\u002Fship verb (`ship` \u002F `build` \u002F `implement`) | `\u002Fautospec-project ship \u003Curl>` (no gate) |\n| GitHub Projects board URL alone, **no** build\u002Fship verb | `\u002Fautospec-project \u003Curl>` (bare resolve — zero mutation, no gate) |\n\n**GitHub Projects board routing (Task 12).** A message that contains a GitHub\nProjects v2 board URL — `https:\u002F\u002Fgithub.com\u002Forgs\u002F\u003Corg>\u002Fprojects\u002F\u003Cn>` or\n`https:\u002F\u002Fgithub.com\u002Fusers\u002F\u003Cuser>\u002Fprojects\u002F\u003Cn>`, optionally with a trailing\n`\u002Fviews\u002F\u003Cn>` — carries board intent, checked before the generic\nimplement\u002Fbuild\u002Fship and `autospec …` rows above so it wins over a bare\n`autospec` or `ship` word in the same message. The classifier reports this as\nskill `autospec-project` with trigger `project-ship` or `project-resolve`;\nextract the URL from the user's message with the same regex and build the\ncommand yourself, since the classifier does not echo the URL back:\n\n- **URL + ship verb** (e.g. \"autospec ship this project for me:\n  `https:\u002F\u002Fgithub.com\u002Forgs\u002FInferWeave\u002Fprojects\u002F2`\") → invoke\n  `\u002Fautospec-project ship \u003Curl>`. `ship` starts autonomous multi-repo work,\n  so this requires the verb explicitly present — a bare URL never implies it.\n- **URL alone**, no ship verb (e.g. a pasted board link, or \"what's on\n  `\u003Curl>`\") → invoke `\u002Fautospec-project \u003Curl>` (bare mode: resolves and\n  prints the plan, performs zero mutation). Reading a board is not a\n  commitment to drain it — this asymmetry is deliberate.\n\nBoth rows still honor the intent gate (biased to false-negatives) and the\none-line opt-out below.\n\n**Auto-route behavior.** When the classifier returns `match:true` with `intent:imperative`, print exactly one line then invoke the mapped skill via the harness skill-invocation primitive:\n\n```\nRouting to \u002F\u003Cskill> — say \"plain\" to opt out.\n```\n\nIf `gate` is non-null, perform the gate check first and skip routing (plain mode) if unmet. For `gate:explore-confirm` (skill `autospec-explore`), do NOT auto-route on the one-line opt-out alone: first print exactly that `\u002Fautospec-explore` starts a perpetual autonomous research + ship loop on an isolated sandbox branch whose PRs target the sandbox branch, never main, then require ONE explicit confirmation and invoke `\u002Fautospec-explore` only on an affirmative reply; the existing `plain`\u002F`no` escape still cancels. If `chain` is non-null, after the mapped skill completes, invoke the chained skill (printing one opt-out line as usual).\n\nOn a non-match or `intent:incidental`\u002F`none`, do nothing — the session proceeds normally in plain mode.\n\n**Post-approval execution-ready routing.** After an autospec\u002Fspec\u002Frefine flow writes\nan approved spec or implementation plan, do not ask the user to choose between\nsubagent-driven and inline execution. Record the recent workflow state for the\nclassifier using the narrow env hints below, then pass the user's approval turn\nverbatim:\n\n- `AUTOSPEC_LISTENER_APPROVED_SPEC_PATH` — path to the approved spec\u002Fdesign.\n- `AUTOSPEC_LISTENER_PLAN_PATH` — path to the written implementation plan.\n- `AUTOSPEC_LISTENER_SOURCE_SPEC_PATH` — source spec path, if the plan was derived from one.\n- `AUTOSPEC_LISTENER_REPO` \u002F `AUTOSPEC_LISTENER_BRANCH` \u002F `AUTOSPEC_LISTENER_BASE` — minimal repo routing metadata.\n- `AUTOSPEC_LISTENER_PLAN_TS` — timestamp for the saved plan or handoff artifact.\n- `AUTOSPEC_LISTENER_AUTOSPEC_REQUESTED=1` — the thread previously requested autospec routing.\n- `AUTOSPEC_LISTENER_AUTONOMOUS_REQUESTED=1` — the thread previously requested autonomous routing.\n- `AUTOSPEC_LISTENER_AUTO_IMPLEMENT_OPEN=\u003Ccount>` — open `auto-implement` issue count, from `gh issue list --repo \u003Crepo> --label auto-implement --state open --json number --jq 'length'`.\n\nFor approval phrases such as `looks good` or `approved`, the classifier emits\n`trigger:\"post_approval_execution_ready\"`. If\n`AUTOSPEC_LISTENER_AUTO_IMPLEMENT_OPEN` is greater than zero, route to\n`\u002Fautospec-run`; otherwise route to `\u002Fautospec --autonomous`. Print exactly:\n\n```\nRouting to \u002Fautospec --autonomous — say \"plain\" to opt out.\n```\n\nor, when open `auto-implement` issues exist:\n\n```\nRouting to \u002Fautospec-run — say \"plain\" to opt out.\n```\n\nExplicit opt-outs (`plain`, `no`, `no workflow`, `just chat`, `do not implement`),\nconceptual questions, destructive\u002Fproduction requests, and descriptive mentions\nof autospec remain plain mode.\n\n**Completed Plan-mode handoff routing.** When any harness exits Plan mode after\nwriting a concrete implementation plan, autospec is the default next step. Set\n`AUTOSPEC_LISTENER_PLAN_EXIT_READY=1` plus\n`AUTOSPEC_LISTENER_PLAN_PATH=\u003Cpath>` (or\n`AUTOSPEC_LISTENER_PLAN_ARTIFACT=\u003Csummary>` when no path exists) and pass the\nlatest Plan-mode exit text or the user's approval turn through the classifier.\nThe classifier emits `trigger:\"plan_exit_ready\"` for the direct Plan-exit\nhandoff and `trigger:\"plan_approved_ready\"` for approval phrases such as\n`continue`, `proceed`, `go`, or `ship it` after a completed plan.\n\nWhen `AUTOSPEC_LISTENER_AUTO_IMPLEMENT_OPEN` is greater than zero, route to\n`\u002Fautospec-run`; otherwise route to `\u002Fautospec --autonomous`. The JSON output\nalso carries `plan_path`, `source_spec_path`, `repo`, `branch`, `base`, and\n`plan_ts` fields so the handoff can pass or summarize the saved plan instead of\nrestarting discovery from scratch. Suppress generic execution-choice prompts\nsuch as \"Subagent-Driven or Inline Execution?\" when this routing state is\npresent; print only the one-line autospec opt-out:\n\n```\nRouting to \u002Fautospec --autonomous — say \"plain\" to opt out.\n```\n\nor:\n\n```\nRouting to \u002Fautospec-run — say \"plain\" to opt out.\n```\n\nDo not route when `AUTOSPEC_LISTENER_PLAN_EXIT_BLOCKED=1`,\n`AUTOSPEC_LISTENER_PLAN_DESTRUCTIVE=1`, the plan text says it is blocked,\nambiguous, destructive, production\u002Fcredential-gated, or still requires explicit\napproval, or the user asks a conceptual question about the plan. The one-turn\nescape includes `plain`, `no`, `no workflow`, `just chat`, `do not implement`,\n`stop`, and `cancel`.\n\n**Escape (D5).** If the user's next turn is the stop word (`plain` or `no`), cancel routing and proceed in plain mode for that request. The existing `file an issue` \u002F `write a spec` triggers continue to fire their own flows below and are unaffected.\n\n## Issue trigger flow\n\nWhen the user utters a phrase from the **Issue triggers** list (see `references\u002Ftrigger-keywords.md`), run this 7-step procedure:\n\n1. **Read recent context.** Pull the last ~10 conversation turns from the harness's message buffer. If that capability is missing, ask the user to paste the relevant excerpt.\n2. **Synthesize draft body.** Build a markdown body with three sections:\n   - `## Goal` — exactly one sentence summarizing what the issue should achieve.\n   - `## Context` — relevant chat excerpts that motivate the issue (verbatim quotes, ≤6 lines).\n   - `## Suggested AC` — ≤3 bulleted acceptance-criteria checkboxes (`- [ ]` style).\n   The full body MUST be ≤400 words.\n3. **Print the draft inline** so the user can read it before approving.\n4. **Confirm via the harness primitive.** Ask exactly: `File this as a new issue? [yes \u002F edit \u002F cancel]`. Use `AskUserQuestion` on Claude Code, an inline prompt elsewhere; if no question primitive is available, ask in the response and wait for the next turn.\n5. **On `yes`**, infer the title and run:\n   ```bash\n   issue_url=\"$(gh issue create --title \"\u003Cinferred-title>\" --body \"\u003Cdraft-body>\" --label needs-classify)\"\n   bash \"${AUTOSPEC_SCRIPTS_DIR:-$HOME\u002F.autospec\u002Fscripts}\u002Fproject-sync-issue.sh\" \"$issue_url\" \"$PWD\"\n   ```\n   Then post the resulting issue URL back to the chat. Project sync is best-effort but journals the issue before remote reconciliation, so a failed projection remains retryable. The `needs-classify` label keeps the issue out of the auto-implement queue until `\u002Fautospec-classify` sizes it.\n6. **On `edit`**, accept free-text revisions from the user, re-render the draft, and re-confirm (loop back to step 4).\n7. **On `cancel`**, print `OK, cancelled.` and silently exit. No persistence, no learning, no suppression list.\n\n### Title-inference rule\n\nBuild the title from the Goal sentence using the following deterministic rule:\n\n1. Take the `## Goal` sentence verbatim.\n2. **Strip leading verbs** `Implement` \u002F `Add` \u002F `Fix` (case-insensitive, with any trailing whitespace) from the front of the sentence.\n3. **Truncate** the remainder to 80 characters total (after the prefix in step 4 is added).\n4. **Prefix with a conventional-commit tag** inferred from the original leading verb:\n   - `Add` \u002F `Implement` \u002F new behavior → `feat: `\n   - `Fix` \u002F bug language → `fix: `\n   - documentation-only language → `docs: `\n   - refactoring language → `refactor: `\n   If inference is ambiguous, leave the tag off entirely and let `\u002Fautospec-classify` add one later.\n\nExample: Goal `Implement a \u002Floop slash command that re-runs the last skill` → title `feat: a \u002Floop slash command that re-runs the last skill`.\n\n## Spec trigger flow\n\nWhen the user utters a phrase from the **Spec triggers** list (see `references\u002Ftrigger-keywords.md`), this listener acts as a **router only** — it does NOT duplicate the Phase 2 brainstorming logic that lives in `\u002Fautospec-define`.\n\n1. **Confirm.** Ask exactly: `Start the autospec design Q&A on \u003Cinferred-topic>? [yes \u002F cancel]`. Use `AskUserQuestion` on Claude Code, an inline prompt elsewhere; if no question primitive is available, ask in the response and wait for the next turn. The `\u003Cinferred-topic>` is a short noun phrase synthesized from the trigger phrase and the most recent user turn (≤10 words).\n2. **On `yes`**, hand off to `\u002Fautospec-define` via the harness's skill-invocation primitive (Claude Code: `Skill` tool; OpenCode: nested skill invocation; Codex CLI: emit the `\u002Fautospec-define` slash command). The downstream skill applies its own tier policy. Do NOT inline any Phase 2 questions here.\n3. **On `cancel`**, print `OK, cancelled.` and silently exit. No persistence, no learning, no suppression list.\n\n## Autonomous mode\n\nThe deterministic listener-match classifier recognizes autonomous-intent\nphrasing and routes the request with `--autonomous` so downstream\n`\u002Fautospec`, `\u002Fautospec-define`, and `\u002Fautospec-run` skip user-facing\nconfirmation gates.\n\nTrigger phrases (case-insensitive, classified imperative):\n\n- \"fix X automatically\"\n- \"no confirmation\"\n- \"just do it\"\n- \"non-interactive\"\n- \"go autonomous\"\n- \"autonomous mode\"\n- \"run autonomously\"\n\nOn a `match:true` + `intent:imperative` classification where the message\ncontains one of these phrases, the listener routes to the mapped autospec\nskill with `--autonomous` appended:\n\n```\nRouting to \u002Fautospec --autonomous — say \"plain\" to opt out.\n```\n\nIf the classifier returns `intent:incidental` or `none`, do NOT route — the\nsession continues in plain mode. The same biased-to-false-negative rule\napplies: descriptive, past-tense, negated, and interrogative uses MUST NOT\nroute.\n\nThe autonomous routing inherits the listener's existing `plain`\u002F`no` escape:\nthe user's next-turn stop word cancels routing and proceeds in plain mode\nfor that request.\n","# autospec-listen (harness-neutral)\n\nConversation listener that watches for issue-filing or spec-writing intent\nmid-conversation and dispatches to the right downstream flow without forcing\nthe user to remember a slash command.\n\n\u003C!-- autospec-block:startup-self-update SKILL_NAME=autospec-listen -->\n\n## Self-update mode\n\nIf the feature-request argument matches the regex `^\\s*update\\s*$` (case-insensitive, whitespace-padded), this skill enters self-update mode and does not run the normal trigger flow:\n\n1. **Detect harness** by checking which install path exists for this skill:\n   - Claude Code: `~\u002F.claude\u002Fskills\u002Fautospec-listen\u002FSKILL.md`\n   - OpenCode:    `~\u002F.config\u002Fopencode\u002Fagent\u002Fautospec-listen.md`\n   - Codex CLI:   `~\u002F.codex\u002Fprompts\u002Fautospec-listen.md`\n2. **Re-install the full autospec suite from `main`** by piping the canonical installer:\n   ```bash\n   curl -fsSL https:\u002F\u002Fraw.githubusercontent.com\u002Fberlinguyinca\u002Fautospec\u002Fmain\u002Fbootstrap.sh | bash -s -- --skill all --harness all --update\n   ```\n   Run this one-liner once; it refreshes all autospec skills across all harnesses.\n3. **Show the diff** between the prior installed file(s) and the freshly fetched copy (e.g. `diff \u003C(cat \u003Cprior>) \u003C(curl -fsSL ...SKILL.md)` or the equivalent recorded by the installer).\n4. **Stop.** Do not enter any trigger flow. Print the upgrade summary and return to the user.\n\nIf no install path is detected, print `Self-update: no installed copy of autospec-listen found; run install.sh first.` and exit.\n\n## Harness detection (run once at skill start)\n\nDetect your harness by checking available tools before dispatching work:\n\n1. **Claude Code** — the `Agent` tool with a `subagent_type` parameter is available.\n   - `TIER_A` = `opus` + `ultrathink`  (the alias, which resolves to the current Claude Opus generation)\n   - `TIER_B` = `sonnet`               (the alias, which resolves to the current Claude Sonnet generation)\n\n2. **OpenCode** — a `task` tool with model\u002Ftier configuration is available (no `subagent_type`).\n   - `TIER_A` = top-tier task model + high reasoning\n   - `TIER_B` = smaller-tier task model + medium reasoning\n\n3. **Codex CLI** — neither `Agent` nor a configurable `task` tool is available; `apply_patch` is the primary edit tool.\n   - `TIER_A` = current top GPT model + `reasoning_effort=high`\n   - `TIER_B` = the spark \u002F cost-optimized variant of the configured model when one exists, else the configured model + `reasoning_effort=medium`\n\n**Fallback rule:** If `TIER_B` is not available in your harness (model unknown, quota\u002Fcapacity failure, authorization failure, or tool call returns an error for that model), silently retry the same dispatch with `TIER_A`. Preserve the parent context on retry; for Codex native subagents, fork\u002Finherit the current conversation context and use the latest top GPT model instead of moving the work into the main session. Never ask the user.\n\nHold `TIER_A` and `TIER_B` for the entire skill run.\n\n## Required capabilities & harness adapter\n\nThis skill assumes four capabilities. Map each one to your harness's actual tool. If a capability is missing, use the listed fallback.\n\n| Capability                  | Claude Code                          | OpenCode                                 | Codex CLI                                | Fallback if missing                                |\n|-----------------------------|--------------------------------------|------------------------------------------|------------------------------------------|----------------------------------------------------|\n| Read recent conversation    | harness-native message buffer        | harness-native message buffer            | harness-native message buffer            | Ask the user to paste the relevant turns           |\n| Ask the user a question     | `AskUserQuestion`                    | inline prompt                            | inline prompt                            | Ask in the response and wait for the next turn     |\n| Run `gh` CLI                | shell tool                           | shell tool                               | shell tool                               | Have the user run the command and confirm          |\n| Hand off to another skill   | `Skill` invocation (`\u002Fautospec-define`) | nested skill invocation                | `\u002Fautospec-define` slash command         | Print the suggested command for the user to run    |\n| Subagent model tier         | Tier B: `sonnet` + medium thinking; Tier A only if escalating to \u002Fautospec-define which honors its own tier policy | Tier B: smaller-tier `task` + medium reasoning; escalation honors target skill | Tier B: current spark\u002Fcost-optimized Codex + `reasoning_effort=medium`; escalation honors target | Honor AGENTS.md tier mapping; retry the same subagent UP on unavailability |\n\u003C!-- autospec-block:harness-adapter-core -->\n\n**Persistent project notes**: write durable preferences to **`AGENTS.md`** in the repo root — this is the de-facto standard recognized by Claude Code (also reads `CLAUDE.md`), OpenCode, and Codex. Per AGENTS.md, this listener runs at **Tier B** (it's a router, not a designer); when it hands off to `\u002Fautospec-define` for a spec trigger, the downstream skill applies its own tier policy. Fall back UP the tier on quota\u002Fcapacity or other unavailability by retrying the same subagent with the stronger tier while preserving parent context.\n\n## Trigger flow\n\n> **Model tier:** Tier B (implementation work) — this listener is a router; subagent dispatches inside its trigger flows (drafting issue bodies, summarizing recent conversation) use cheaper models per AGENTS.md. Spec-trigger handoff to `\u002Fautospec-define` lets the downstream skill apply its own tier policy.\n\n## Keyword auto-routing\n\nBeyond the explicit issue\u002Fspec triggers, this listener routes common build\u002Fchange verbs into the matching autospec skill so the pipeline is the default path. Routing is gated by an imperative-intent check and always offers a one-line opt-out.\n\n**Intent gate.** Do not classify by eye. Delegate to the deterministic classifier, passing the user's latest message verbatim:\n\n```bash\nbash \"${AUTOSPEC_SCRIPTS_DIR:-$HOME\u002F.autospec\u002Fscripts}\u002Flistener-match.sh\" --classify \"\u003Cuser message>\"\n```\n\nIt emits `{\"match\":bool,\"skill\":...,\"trigger\":...,\"intent\":\"imperative|incidental|none\",\"confidence\":0-1,\"autonomous\":bool,\"chain\":\u003Cskill|null>,\"gate\":\u003Cstring|null>}`. The gate is biased to false-negatives: descriptive (\"the design is nice\"), past-tense (\"I reviewed it\"), negated (\"don't implement yet\"), and interrogative (\"should we redesign?\") uses return `match:false` and MUST NOT route. Prefer no route over a misfire.\n\nThe two new trailing fields:\n- `chain` — after the primary skill completes, invoke the chained skill (e.g. `\u002Fautospec-run` after `\u002Fautospec-refine`). Null when no chain applies.\n- `gate` — before routing, the listener must satisfy the named context gate; `auto-implement-open` means run `gh issue list --repo \u003Crepo> --label auto-implement --state open --json number --jq 'length'` and route only if > 0, else stay in plain mode. `explore-confirm` means the route launches `\u002Fautospec-explore`, which starts a perpetual autonomous research + ship loop on an isolated sandbox branch whose PRs target the sandbox branch, never main; before invoking, print exactly that consequence and require ONE explicit confirmation, routing only on an affirmative reply. Null means no gate.\n\n**Verb → skill map (D3).**\n\n| Verb(s) in an imperative request | Routes to |\n|----------------------------------|-----------|\n| `design` \u002F `new feature` \u002F `spec` | `\u002Fautospec-define` |\n| `implement` \u002F `build` \u002F `ship` | `\u002Fautospec-run` (or `\u002Fautospec` end-to-end when no issues exist yet) |\n| `refine` \u002F `optimize` \u002F `polish` \u002F `improve` \u002F `tune` | `\u002Fautospec-refine` |\n| `run` (scoped: \"run it\", \"run the loop\", \"run autospec\"; bare \"run\" only when open `auto-implement` issues exist) | `\u002Fautospec-run` |\n| `refine and run` \u002F `tune up` (combined) | `\u002Fautospec-refine` then `\u002Fautospec-run` |\n| `review` | `\u002Fautospec-review` |\n| `explore` \u002F `discover` (only with an explicit build\u002Fship\u002Fimprove action connector, e.g. \"explore and ship\", \"discover features to build\") | `\u002Fautospec-explore` (gate `explore-confirm`) |\n| `fix` (imperative, e.g. \"fix the login flow\"; trivial-fix phrasings like \"fix the typo\", \"quick fix\", \"fix the formatting\" stay plain) | the umbrella `\u002Fautospec` (no gate) |\n| `autospec …` | the umbrella `\u002Fautospec` |\n| GitHub Projects board URL (`https:\u002F\u002Fgithub.com\u002F(orgs\\|users)\u002F\u003Cname>\u002Fprojects\u002F\u003Cn>`, optional `\u002Fviews\u002F\u003Cn>`) **with** a build\u002Fship verb (`ship` \u002F `build` \u002F `implement`) | `\u002Fautospec-project ship \u003Curl>` (no gate) |\n| GitHub Projects board URL alone, **no** build\u002Fship verb | `\u002Fautospec-project \u003Curl>` (bare resolve — zero mutation, no gate) |\n\n**GitHub Projects board routing (Task 12).** A message that contains a GitHub\nProjects v2 board URL — `https:\u002F\u002Fgithub.com\u002Forgs\u002F\u003Corg>\u002Fprojects\u002F\u003Cn>` or\n`https:\u002F\u002Fgithub.com\u002Fusers\u002F\u003Cuser>\u002Fprojects\u002F\u003Cn>`, optionally with a trailing\n`\u002Fviews\u002F\u003Cn>` — carries board intent, checked before the generic\nimplement\u002Fbuild\u002Fship and `autospec …` rows above so it wins over a bare\n`autospec` or `ship` word in the same message. The classifier reports this as\nskill `autospec-project` with trigger `project-ship` or `project-resolve`;\nextract the URL from the user's message with the same regex and build the\ncommand yourself, since the classifier does not echo the URL back:\n\n- **URL + ship verb** (e.g. \"autospec ship this project for me:\n  `https:\u002F\u002Fgithub.com\u002Forgs\u002FInferWeave\u002Fprojects\u002F2`\") → invoke\n  `\u002Fautospec-project ship \u003Curl>`. `ship` starts autonomous multi-repo work,\n  so this requires the verb explicitly present — a bare URL never implies it.\n- **URL alone**, no ship verb (e.g. a pasted board link, or \"what's on\n  `\u003Curl>`\") → invoke `\u002Fautospec-project \u003Curl>` (bare mode: resolves and\n  prints the plan, performs zero mutation). Reading a board is not a\n  commitment to drain it — this asymmetry is deliberate.\n\nBoth rows still honor the intent gate (biased to false-negatives) and the\none-line opt-out below.\n\n**Auto-route behavior.** When the classifier returns `match:true` with `intent:imperative`, print exactly one line then invoke the mapped skill via the harness skill-invocation primitive:\n\n```\nRouting to \u002F\u003Cskill> — say \"plain\" to opt out.\n```\n\nIf `gate` is non-null, perform the gate check first and skip routing (plain mode) if unmet. For `gate:explore-confirm` (skill `autospec-explore`), do NOT auto-route on the one-line opt-out alone: first print exactly that `\u002Fautospec-explore` starts a perpetual autonomous research + ship loop on an isolated sandbox branch whose PRs target the sandbox branch, never main, then require ONE explicit confirmation and invoke `\u002Fautospec-explore` only on an affirmative reply; the existing `plain`\u002F`no` escape still cancels. If `chain` is non-null, after the mapped skill completes, invoke the chained skill (printing one opt-out line as usual).\n\nOn a non-match or `intent:incidental`\u002F`none`, do nothing — the session proceeds normally in plain mode.\n\n**Post-approval execution-ready routing.** After an autospec\u002Fspec\u002Frefine flow writes\nan approved spec or implementation plan, do not ask the user to choose between\nsubagent-driven and inline execution. Record the recent workflow state for the\nclassifier using the narrow env hints below, then pass the user's approval turn\nverbatim:\n\n- `AUTOSPEC_LISTENER_APPROVED_SPEC_PATH` — path to the approved spec\u002Fdesign.\n- `AUTOSPEC_LISTENER_PLAN_PATH` — path to the written implementation plan.\n- `AUTOSPEC_LISTENER_SOURCE_SPEC_PATH` — source spec path, if the plan was derived from one.\n- `AUTOSPEC_LISTENER_REPO` \u002F `AUTOSPEC_LISTENER_BRANCH` \u002F `AUTOSPEC_LISTENER_BASE` — minimal repo routing metadata.\n- `AUTOSPEC_LISTENER_PLAN_TS` — timestamp for the saved plan or handoff artifact.\n- `AUTOSPEC_LISTENER_AUTOSPEC_REQUESTED=1` — the thread previously requested autospec routing.\n- `AUTOSPEC_LISTENER_AUTONOMOUS_REQUESTED=1` — the thread previously requested autonomous routing.\n- `AUTOSPEC_LISTENER_AUTO_IMPLEMENT_OPEN=\u003Ccount>` — open `auto-implement` issue count, from `gh issue list --repo \u003Crepo> --label auto-implement --state open --json number --jq 'length'`.\n\nFor approval phrases such as `looks good` or `approved`, the classifier emits\n`trigger:\"post_approval_execution_ready\"`. If\n`AUTOSPEC_LISTENER_AUTO_IMPLEMENT_OPEN` is greater than zero, route to\n`\u002Fautospec-run`; otherwise route to `\u002Fautospec --autonomous`. Print exactly:\n\n```\nRouting to \u002Fautospec --autonomous — say \"plain\" to opt out.\n```\n\nor, when open `auto-implement` issues exist:\n\n```\nRouting to \u002Fautospec-run — say \"plain\" to opt out.\n```\n\nExplicit opt-outs (`plain`, `no`, `no workflow`, `just chat`, `do not implement`),\nconceptual questions, destructive\u002Fproduction requests, and descriptive mentions\nof autospec remain plain mode.\n\n**Completed Plan-mode handoff routing.** When any harness exits Plan mode after\nwriting a concrete implementation plan, autospec is the default next step. Set\n`AUTOSPEC_LISTENER_PLAN_EXIT_READY=1` plus\n`AUTOSPEC_LISTENER_PLAN_PATH=\u003Cpath>` (or\n`AUTOSPEC_LISTENER_PLAN_ARTIFACT=\u003Csummary>` when no path exists) and pass the\nlatest Plan-mode exit text or the user's approval turn through the classifier.\nThe classifier emits `trigger:\"plan_exit_ready\"` for the direct Plan-exit\nhandoff and `trigger:\"plan_approved_ready\"` for approval phrases such as\n`continue`, `proceed`, `go`, or `ship it` after a completed plan.\n\nWhen `AUTOSPEC_LISTENER_AUTO_IMPLEMENT_OPEN` is greater than zero, route to\n`\u002Fautospec-run`; otherwise route to `\u002Fautospec --autonomous`. The JSON output\nalso carries `plan_path`, `source_spec_path`, `repo`, `branch`, `base`, and\n`plan_ts` fields so the handoff can pass or summarize the saved plan instead of\nrestarting discovery from scratch. Suppress generic execution-choice prompts\nsuch as \"Subagent-Driven or Inline Execution?\" when this routing state is\npresent; print only the one-line autospec opt-out:\n\n```\nRouting to \u002Fautospec --autonomous — say \"plain\" to opt out.\n```\n\nor:\n\n```\nRouting to \u002Fautospec-run — say \"plain\" to opt out.\n```\n\nDo not route when `AUTOSPEC_LISTENER_PLAN_EXIT_BLOCKED=1`,\n`AUTOSPEC_LISTENER_PLAN_DESTRUCTIVE=1`, the plan text says it is blocked,\nambiguous, destructive, production\u002Fcredential-gated, or still requires explicit\napproval, or the user asks a conceptual question about the plan. The one-turn\nescape includes `plain`, `no`, `no workflow`, `just chat`, `do not implement`,\n`stop`, and `cancel`.\n\n**Escape (D5).** If the user's next turn is the stop word (`plain` or `no`), cancel routing and proceed in plain mode for that request. The existing `file an issue` \u002F `write a spec` triggers continue to fire their own flows below and are unaffected.\n\n## Issue trigger flow\n\nWhen the user utters a phrase from the **Issue triggers** list (see `references\u002Ftrigger-keywords.md`), run this 7-step procedure:\n\n1. **Read recent context.** Pull the last ~10 conversation turns from the harness's message buffer. If that capability is missing, ask the user to paste the relevant excerpt.\n2. **Synthesize draft body.** Build a markdown body with three sections:\n   - `## Goal` — exactly one sentence summarizing what the issue should achieve.\n   - `## Context` — relevant chat excerpts that motivate the issue (verbatim quotes, ≤6 lines).\n   - `## Suggested AC` — ≤3 bulleted acceptance-criteria checkboxes (`- [ ]` style).\n   The full body MUST be ≤400 words.\n3. **Print the draft inline** so the user can read it before approving.\n4. **Confirm via the harness primitive.** Ask exactly: `File this as a new issue? [yes \u002F edit \u002F cancel]`. Use `AskUserQuestion` on Claude Code, an inline prompt elsewhere; if no question primitive is available, ask in the response and wait for the next turn.\n5. **On `yes`**, infer the title and run:\n   ```bash\n   issue_url=\"$(gh issue create --title \"\u003Cinferred-title>\" --body \"\u003Cdraft-body>\" --label needs-classify)\"\n   bash \"${AUTOSPEC_SCRIPTS_DIR:-$HOME\u002F.autospec\u002Fscripts}\u002Fproject-sync-issue.sh\" \"$issue_url\" \"$PWD\"\n   ```\n   Then post the resulting issue URL back to the chat. Project sync is best-effort but journals the issue before remote reconciliation, so a failed projection remains retryable. The `needs-classify` label keeps the issue out of the auto-implement queue until `\u002Fautospec-classify` sizes it.\n6. **On `edit`**, accept free-text revisions from the user, re-render the draft, and re-confirm (loop back to step 4).\n7. **On `cancel`**, print `OK, cancelled.` and silently exit. No persistence, no learning, no suppression list.\n\n### Title-inference rule\n\nBuild the title from the Goal sentence using the following deterministic rule:\n\n1. Take the `## Goal` sentence verbatim.\n2. **Strip leading verbs** `Implement` \u002F `Add` \u002F `Fix` (case-insensitive, with any trailing whitespace) from the front of the sentence.\n3. **Truncate** the remainder to 80 characters total (after the prefix in step 4 is added).\n4. **Prefix with a conventional-commit tag** inferred from the original leading verb:\n   - `Add` \u002F `Implement` \u002F new behavior → `feat: `\n   - `Fix` \u002F bug language → `fix: `\n   - documentation-only language → `docs: `\n   - refactoring language → `refactor: `\n   If inference is ambiguous, leave the tag off entirely and let `\u002Fautospec-classify` add one later.\n\nExample: Goal `Implement a \u002Floop slash command that re-runs the last skill` → title `feat: a \u002Floop slash command that re-runs the last skill`.\n\n## Spec trigger flow\n\nWhen the user utters a phrase from the **Spec triggers** list (see `references\u002Ftrigger-keywords.md`), this listener acts as a **router only** — it does NOT duplicate the Phase 2 brainstorming logic that lives in `\u002Fautospec-define`.\n\n1. **Confirm.** Ask exactly: `Start the autospec design Q&A on \u003Cinferred-topic>? [yes \u002F cancel]`. Use `AskUserQuestion` on Claude Code, an inline prompt elsewhere; if no question primitive is available, ask in the response and wait for the next turn. The `\u003Cinferred-topic>` is a short noun phrase synthesized from the trigger phrase and the most recent user turn (≤10 words).\n2. **On `yes`**, hand off to `\u002Fautospec-define` via the harness's skill-invocation primitive (Claude Code: `Skill` tool; OpenCode: nested skill invocation; Codex CLI: emit the `\u002Fautospec-define` slash command). The downstream skill applies its own tier policy. Do NOT inline any Phase 2 questions here.\n3. **On `cancel`**, print `OK, cancelled.` and silently exit. No persistence, no learning, no suppression list.\n\n## Autonomous mode\n\nThe deterministic listener-match classifier recognizes autonomous-intent\nphrasing and routes the request with `--autonomous` so downstream\n`\u002Fautospec`, `\u002Fautospec-define`, and `\u002Fautospec-run` skip user-facing\nconfirmation gates.\n\nTrigger phrases (case-insensitive, classified imperative):\n\n- \"fix X automatically\"\n- \"no confirmation\"\n- \"just do it\"\n- \"non-interactive\"\n- \"go autonomous\"\n- \"autonomous mode\"\n- \"run autonomously\"\n\nOn a `match:true` + `intent:imperative` classification where the message\ncontains one of these phrases, the listener routes to the mapped autospec\nskill with `--autonomous` appended:\n\n```\nRouting to \u002Fautospec --autonomous — say \"plain\" to opt out.\n```\n\nIf the classifier returns `intent:incidental` or `none`, do NOT route — the\nsession continues in plain mode. The same biased-to-false-negative rule\napplies: descriptive, past-tense, negated, and interrogative uses MUST NOT\nroute.\n\nThe autonomous routing inherits the listener's existing `plain`\u002F`no` escape:\nthe user's next-turn stop word cancels routing and proceeds in plain mode\nfor that request.\n",{"name":5,"description":7},"main","9e7f04357ff46ab731d4f56a4ac5ca5449c13bb0","skills\u002Fautospec-listen",1,"https:\u002F\u002Fgithub.com\u002Fberlinguyinca\u002Fautospec","https:\u002F\u002Fgithub.com\u002Fberlinguyinca","Love to lead, code, cook and contribute to open source and believe that you can learn anything you set your mind too.\r\n","2026-11-11 07:09:35",{"source":8,"skillName":5,"commands":58,"prompt":68,"zipAvailable":69},[59,62,65],{"tool":60,"command":61},"npx","npx skills add berlinguyinca\u002Fautospec --skill autospec-listen",{"tool":63,"command":64},"bunx","bunx skills add berlinguyinca\u002Fautospec --skill autospec-listen",{"tool":66,"command":67},"pnpm","pnpm dlx skills add berlinguyinca\u002Fautospec --skill autospec-listen","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fberlinguyinca\u002Fautospec\u002Fblob\u002Fmain\u002Fskills\u002Fautospec-listen\u002FSKILL.md",true,[71,77,83,89,92,96,100,103,106,110],{"path":72,"resourceType":73,"resourceTypeDesc":74,"sizeBytes":75,"sourceHtmlUrl":76},"codex\u002F.gitkeep","OTHER","其他",0,"https:\u002F\u002Fgithub.com\u002Fberlinguyinca\u002Fautospec\u002Fblob\u002Fmain\u002Fskills\u002Fautospec-listen\u002Fcodex\u002F.gitkeep",{"path":78,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":81,"sourceHtmlUrl":82},"codex\u002Fprompt.md","REFERENCE","参考文档",19746,"https:\u002F\u002Fgithub.com\u002Fberlinguyinca\u002Fautospec\u002Fblob\u002Fmain\u002Fskills\u002Fautospec-listen\u002Fcodex\u002Fprompt.md",{"path":84,"resourceType":85,"resourceTypeDesc":86,"sizeBytes":87,"sourceHtmlUrl":88},"install.sh","SCRIPT","脚本",13379,"https:\u002F\u002Fgithub.com\u002Fberlinguyinca\u002Fautospec\u002Fblob\u002Fmain\u002Fskills\u002Fautospec-listen\u002Finstall.sh",{"path":90,"resourceType":73,"resourceTypeDesc":74,"sizeBytes":75,"sourceHtmlUrl":91},"opencode\u002F.gitkeep","https:\u002F\u002Fgithub.com\u002Fberlinguyinca\u002Fautospec\u002Fblob\u002Fmain\u002Fskills\u002Fautospec-listen\u002Fopencode\u002F.gitkeep",{"path":93,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":94,"sourceHtmlUrl":95},"opencode\u002Fagent.md",20591,"https:\u002F\u002Fgithub.com\u002Fberlinguyinca\u002Fautospec\u002Fblob\u002Fmain\u002Fskills\u002Fautospec-listen\u002Fopencode\u002Fagent.md",{"path":97,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":98,"sourceHtmlUrl":99},"README.md",2675,"https:\u002F\u002Fgithub.com\u002Fberlinguyinca\u002Fautospec\u002Fblob\u002Fmain\u002Fskills\u002Fautospec-listen\u002FREADME.md",{"path":101,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":75,"sourceHtmlUrl":102},"references\u002F.gitkeep","https:\u002F\u002Fgithub.com\u002Fberlinguyinca\u002Fautospec\u002Fblob\u002Fmain\u002Fskills\u002Fautospec-listen\u002Freferences\u002F.gitkeep",{"path":104,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":75,"sourceHtmlUrl":105},"references\u002Ffixture-conversations\u002F.gitkeep","https:\u002F\u002Fgithub.com\u002Fberlinguyinca\u002Fautospec\u002Fblob\u002Fmain\u002Fskills\u002Fautospec-listen\u002Freferences\u002Ffixture-conversations\u002F.gitkeep",{"path":107,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":108,"sourceHtmlUrl":109},"references\u002Ftrigger-keywords.md",1079,"https:\u002F\u002Fgithub.com\u002Fberlinguyinca\u002Fautospec\u002Fblob\u002Fmain\u002Fskills\u002Fautospec-listen\u002Freferences\u002Ftrigger-keywords.md",{"path":111,"resourceType":85,"resourceTypeDesc":86,"sizeBytes":112,"sourceHtmlUrl":113},"uninstall.sh",3940,"https:\u002F\u002Fgithub.com\u002Fberlinguyinca\u002Fautospec\u002Fblob\u002Fmain\u002Fskills\u002Fautospec-listen\u002Funinstall.sh",[115,124,133,142,166,190],{"skillNo":116,"name":117,"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":118,"categories":119,"keywords":121,"publishedAt":44},"01M71RJHYMKPTC9AAPXWQ6F4RT","trio-pass-skill","https:\u002F\u002Fgithub.com\u002Fberlinguyinca\u002Fautospec\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Flockstep-trio-pass\u002FSKILL.md",[120],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},[122,123],"trio","trio pass skill",{"skillNo":125,"name":126,"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":127,"categories":128,"keywords":130,"publishedAt":44},"01M71RJFBQ1KAF9XM3T3Y4VMSK","duo-pass-skill","https:\u002F\u002Fgithub.com\u002Fberlinguyinca\u002Fautospec\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Flockstep-duo-pass\u002FSKILL.md",[129],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},[131,132],"duo pass skill","duo",{"skillNo":134,"name":135,"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":136,"categories":137,"keywords":139,"publishedAt":44},"01M71RJDT1TB3P9TRWR6A534FB","duo-divergence-skill","https:\u002F\u002Fgithub.com\u002Fberlinguyinca\u002Fautospec\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Flockstep-duo-divergence\u002FSKILL.md",[138],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},[140,132,141],"duo divergence skill","divergence",{"skillNo":143,"name":24,"nameZh":6,"description":144,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":145,"categories":146,"keywords":150,"publishedAt":44},"01M71RJAC4NNEQSYBDSY3NZS73","Use when the user wants to ship a feature end-to-end across multiple commits — bootstraps repo if missing, brainstorms a design spec, decomposes into linked GitHub issues, or splits an existing spec into issues, then runs an autonomous implementation loop with auto-merge.","https:\u002F\u002Fgithub.com\u002Fberlinguyinca\u002Fautospec\u002Fblob\u002Fmain\u002Fskills\u002Fautospec\u002FSKILL.md",[147],{"categoryNo":148,"categoryName":149,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000015","视觉设计",[24,151,152,153,38,154,35,155,34,156,31,157,158,159,160,161,162,163,164,165],"ship a feature","end","issues","autonomous","loop","across","merge","ship","implementation loop","github issues","feature","commits","repo","missing","use",{"skillNo":167,"name":168,"nameZh":6,"description":169,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":170,"categories":171,"keywords":172,"publishedAt":44},"01M71RJ4F9HT8PT26CKNM9XE1M","autospec-upgrade","Use when the user wants to upgrade an Angular, Next.js, or React project to the latest official versions safely — by locking observable behavior before touching versions, upgrading one major at a time with official codemods, and gating completion on mutation score rather than coverage theater.","https:\u002F\u002Fgithub.com\u002Fberlinguyinca\u002Fautospec\u002Fblob\u002Fmain\u002Fskills\u002Fautospec-upgrade\u002FSKILL.md",[],[173,24,174,175,176,177,178,179,180,34,181,182,183,184,165,185,186,187,188,189],"autospec upgrade","upgrade","mutation","coverage","official","project","time","gating","angular","react","next","at","Next.js","one","with","behavior","latest",{"skillNo":191,"name":192,"nameZh":6,"description":193,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":194,"categories":195,"keywords":201,"publishedAt":44},"01M71RJ28FJ4PM6DXHB7T5CF8H","autospec-ui-audit","Use when a React Router application needs a deterministic whole-site route inventory before UI, UX, navigation, or evidence auditing begins.","https:\u002F\u002Fgithub.com\u002Fberlinguyinca\u002Fautospec\u002Fblob\u002Fmain\u002Fskills\u002Fautospec-ui-audit\u002FSKILL.md",[196],{"categoryNo":197,"categoryName":198,"taxonomyType":199,"taxonomyTypeDesc":200},"01JBSEEDCAT000000000000032","设计师","PROFESSION","适用职业",[202,24,203,204,205,165,206,207,208,209,210,211,212,213,214,182,215,34,216],"ui audit","audit","autospec ui audit","ui","inventory","evidence","deterministic","route","ux","site","react router","navigation","application","needs","router"]