[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7357ENM2WVDGF04Z4R559EY":3,"skill-same:01M7357ENM2WVDGF04Z4R559EY":51},{"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":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":30,"commitSha":31,"skillDirectory":32,"forksCount":33,"repositoryHtmlUrl":34,"repositoryHomepageUrl":35,"ownerHtmlUrl":36,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":37,"install":38,"resources":50},"01M7357ENM2WVDGF04Z4R559EY","purge-hallucinations",null,"Verify and purge fake citations.","Morrison-Lab\u002Fai-config","Morrison-Lab","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F104235520?v=4","01M734G0451S2G35MN4HFQK7XP","Morrison Lab","01M734G059BYFDFQA8YTPDGPMH",4,"MIT","https:\u002F\u002Fgithub.com\u002FMorrison-Lab\u002Fai-config\u002Fblob\u002Fmain\u002Fskills\u002Fpurge-hallucinations\u002FSKILL.md",[],[],"2026-11-11 15:11:14",false,"---\nname: purge-hallucinations\ndescription: \"Verify and purge fake citations.\"\nuser-invocable: true\nallowed-tools:\n  - Bash\n  - Read\n  - Edit\n  - Write\n---\n\n# purge-hallucinations — verify references resolve, then purge the fakes\n\nLLM-written content (memories, skills, `CLAUDE.md`, code, docs, PR\ndescriptions) drifts into **hallucinations**: concrete references that read as\nauthoritative but don't resolve to anything real. This skill checks each\ncheckable reference against ground truth and, for every one proven fabricated,\nproposes a fix interactively.\n\nThe prime directive: **only purge what you can prove is fake.** A reference you\n*can't* verify (network down, ambiguous symbol, private resource) is\n**unverifiable**, not a hallucination — report it, never silently delete it.\n\n## When this fires\n\n- \"purge hallucinations\", \"ph\", \"check for hallucinations\", \"fact-check this\"\n- \"verify the references\", \"find made-up \u002F fabricated references\"\n- \"does everything in this file actually exist\", \"what did you make up\"\n- \"audit my memories\u002Fskills for stale references\"\n- Proactively when reviewing AI-generated text\u002Fcode that asserts concrete,\n  checkable facts (file names, function names, versions, links, citations) —\n  especially memory and instruction files, which compound errors over time.\n\n## Step 1 — Resolve the target\n\nPick the narrowest target the user named, in this precedence order:\n\n1. **Explicit target** — a file, a directory, a PR diff, or pasted AI output\n   the user pointed at. For a PR: `gh pr diff \u003Cn>` (local) or\n   `mcp__github__pull_request_read` `method: get_diff` (remote\u002Fweb — see\n   `memories\u002Fgithub-mcp-tools.md`); audit only the references the diff *introduces*.\n2. **The memory\u002Finstruction corpus** — the ai-config repo's `memories\u002F`,\n   `skills\u002F`, and `CLAUDE.md`. Find the repo root with\n   `${CLAUDE_PLUGIN_ROOT:-$(git -C ~\u002F.claude\u002Fskills\u002Fpurge-hallucinations rev-parse --show-toplevel 2>\u002Fdev\u002Fnull || pwd)}`.\n3. **The current repo's code & docs** — when the user says \"this repo\" or\n   names no target while inside a project.\n\nIf the target is genuinely ambiguous, ask once; otherwise proceed with the\nnarrowest reasonable reading and state which you chose.\n\n## Step 2 — Extract concrete, checkable references\n\nScan the target for references that have a definite right\u002Fwrong answer. Ignore\nprose, opinions, and design rationale — those aren't hallucinations, just\nclaims. Pull out, with file + line for each:\n\n| Reference type | Examples |\n|---|---|\n| File \u002F path | `src\u002Ffoo.R`, `here(\"data\u002Fx.csv\")`, a relative link in a `.qmd` |\n| Function \u002F object \u002F symbol | `pkg::fn()`, an R object, a shell command, a Make target |\n| Action ref + version | `uses: org\u002Faction@v3`, `@\u003Csha>`, a tag\u002Frelease |\n| Skill name | `skills\u002F\u003Cname>\u002F`, `~\u002F.claude\u002Fskills\u002F\u003Cname>\u002F`, a `\u002Fslash-command` |\n| Memory cross-link | `[[some-memory-name]]` in a memory body |\n| URL \u002F link | `https:\u002F\u002F…`, a docs anchor, a badge target |\n| Citation \u002F package | a CRAN\u002FBioconductor package, a DOI, a `DESCRIPTION` dep |\n| Flag \u002F option \u002F config key | a CLI flag, an env var, a YAML key the schema defines |\n| API \u002F SDK method or model id | an SDK method, a model name, an endpoint |\n\n## Step 3 — Verify each against ground truth\n\nUse the cheapest check that *proves existence or absence*. Match the tool to\nthe reference type:\n\n- **File \u002F path** — `test -e \u003Cpath>`, `git ls-files -- \u003Cpath>`, `ls`.\n- **Function \u002F object** — grep for the *definition*, not just a mention\n  (`grep -rn \"fn \u003C- function\" R\u002F`); for an exported R fn check `NAMESPACE` \u002F\n  `Rscript -e 'library(pkg); exists(\"fn\")'`. Shell command — `command -v`.\n- **Action ref + version** — confirm the tag\u002FSHA exists\n  (`gh api repos\u002F\u003Corg>\u002F\u003Crepo>\u002Fgit\u002Frefs\u002Ftags\u002F\u003Ctag>` or `...\u002Fcommits\u002F\u003Csha>`).\n- **Skill name** — `ls skills\u002F\u003Cname>\u002F` in the local ai-config directory. If\n  not found locally, check the session's available-skills list (appears in\n  system reminders) before classifying as ❌; a globally-available system skill\n  with no local directory is ❓ Unverifiable, not ❌ Fabricated.\n- **Memory cross-link** — `[[target]]` links resolve to **skill directories**\n  (`ls skills\u002F\u003Ctarget>\u002F`); if no matching skill, search memory headings\n  (`grep -rn \"^# .*\u003Ctarget>\" memories\u002F`).\n- **URL \u002F link** — `curl -sSI -o \u002Fdev\u002Fnull -w '%{http_code}' \u003Curl>` (local) or\n  `WebFetch` (remote). **A 404\u002F410 is fabricated; a timeout, 403, 429, or\n  DNS failure is *unverifiable*** — distinguish them.\n- **Citation \u002F package** — CRAN: `https:\u002F\u002Fcran.r-project.org\u002Fpackage=\u003Cpkg>`;\n  installed: `Rscript -e 'find.package(\"\u003Cpkg>\")'`; dep: check `DESCRIPTION`;\n  DOI \u002F `.bib`: `python3 scripts\u002Fcheck_doi_bib.py \u003Cpath-to-.bib>` (or `--root \u003Cdir> --cited-only`).\n- **Flag \u002F option \u002F config key** — grep the tool's `--help`, its schema, or\n  its source; for a YAML key, the consuming code\u002Fschema.\n- **API \u002F SDK method or model id** — check the SDK source\u002Fdocs. For\n  Claude\u002FAnthropic specifics (model ids, params), **defer to the `claude-api`\n  skill** rather than guessing.\n\nSort every reference into exactly one bucket:\n\n| Bucket | Meaning | Disposition |\n|---|---|---|\n| ✅ **Resolves** | proven to exist | leave it |\n| ❌ **Fabricated** | proven absent (404, no definition, no such tag\u002Ffile) | propose a fix (Step 4) |\n| ❓ **Unverifiable** | can't be checked here (network, private, ambiguous) | **report, do not edit** |\n\nWhen in doubt, it's ❓ not ❌. False deletions are worse than a flagged\nmaybe.\n\n## Step 4 — Propose a fix for each fabrication (interactive)\n\nDefault action is **propose, then apply on confirmation** — never bulk-edit\nsilently. For each ❌, present: the reference, where it is, the evidence it's\nfake, and a proposed fix — exactly one of:\n\n1. **Correct it** — when a real target is an obvious near-match (typo, moved\n   path, renamed symbol, bumped version). Show the before→after. Prefer this\n   over deletion when a true referent clearly exists (`grep`\u002F`gh` for the\n   nearest real name).\n2. **Remove it** — when the reference is wholly invented and removing it leaves\n   the sentence\u002Fcode correct.\n3. **Rewrite around it** — when removal would leave a dangling clause or broken\n   code; restate without the fake reference.\n\nApply each fix only after the user confirms (batch obvious typo-corrections\ntogether if the user prefers). Re-verify after editing. If the target is a\nmemory or instruction file, the global *\"verify before recommending\"* rule is\nthe reason this matters — a fabricated memory poisons every future session.\n\n## Step 5 — Report\n\nSummarize: N references checked → ✅ resolved, ❌ fixed (list before→after),\n❓ unverifiable (list, with why each couldn't be checked). The ❓ list is a\nfeature, not a failure — it tells the user exactly what still needs a human\neyeball.\n\n## Custom agent for the detect phase\n\nSteps 1--3 (resolve target, extract references, verify against ground truth)\nhave no need for Edit\u002FWrite access. Delegate them to the\n`hallucination-detector` custom agent (`.claude\u002Fagents\u002Fhallucination-detector.md`)\nfor a hard, harness-enforced guarantee against Edit\u002FWrite tool use before the\nreport in Step 4 is reviewed --- tighter than this skill's own\ninstruction-only discipline, though the agent retains `Bash` for read-only\nchecks, so avoiding a write-capable shell command is still instruction-level.\nRun Step 4 (propose\u002Fapply fixes) in the main session afterward.\n\n## Relationship to other skills\n\n- **`record-learnings`, `ums`, `memorize` \u002F `remember`** — these *write* the\n  memory\u002Fskill\u002Finstruction corpus; `purge-hallucinations` *audits* it. A\n  natural hand-off: after a big memory\u002Fskills write, run this to catch\n  fabricated cross-links and stale paths.\n- **`claude-api`** — the source of truth for Claude\u002FAnthropic model ids and\n  params; defer to it instead of guessing when verifying those references.\n- **`simplify` \u002F `tidy`** — clean *structure* (dead code, redundancy); this\n  cleans *truth* (references that don't resolve). Complementary passes.\n- **`reprexes`** — if verifying a code reference needs actually running it,\n  isolate it as a reprex first.\n- **`check-rendered-refs` \u002F `crr`** — the output-side counterpart. This skill\n  audits *source* references; `crr` scans *rendered* HTML for crossrefs\u002Fcitations\n  that broke at render time and leaked into the page as `?@key`. Hand a `?@key`\n  hit here to trace the dangling key back to its source line.\n- **`check-info-quality` \u002F `ciq`** — this skill checks whether a reference\n  **exists**; `ciq`'s check C checks whether an existing, real reference\n  **actually supports** the claim it's attached to. A citation can pass here\n  and still fail `ciq` if it resolves but doesn't say what the text claims.\n\n## Anti-patterns\n\n- ❌ Deleting an **unverifiable** reference as if it were proven fake — the\n  cardinal sin. Network\u002Fpermission failures are ❓, not ❌.\n- ❌ Flagging prose, opinions, or design rationale — only concrete, checkable\n  references are in scope.\n- ❌ Confirming a symbol exists by finding a *mention* of it rather than its\n  *definition*.\n- ❌ Bulk-editing without proposing fixes first (default is interactive).\n- ❌ Treating a network 403\u002F429\u002Ftimeout the same as a 404.\n- ❌ Guessing whether a Claude\u002FAnthropic model id or param is real instead of\n  deferring to `claude-api`.\n","# purge-hallucinations — verify references resolve, then purge the fakes\n\nLLM-written content (memories, skills, `CLAUDE.md`, code, docs, PR\ndescriptions) drifts into **hallucinations**: concrete references that read as\nauthoritative but don't resolve to anything real. This skill checks each\ncheckable reference against ground truth and, for every one proven fabricated,\nproposes a fix interactively.\n\nThe prime directive: **only purge what you can prove is fake.** A reference you\n*can't* verify (network down, ambiguous symbol, private resource) is\n**unverifiable**, not a hallucination — report it, never silently delete it.\n\n## When this fires\n\n- \"purge hallucinations\", \"ph\", \"check for hallucinations\", \"fact-check this\"\n- \"verify the references\", \"find made-up \u002F fabricated references\"\n- \"does everything in this file actually exist\", \"what did you make up\"\n- \"audit my memories\u002Fskills for stale references\"\n- Proactively when reviewing AI-generated text\u002Fcode that asserts concrete,\n  checkable facts (file names, function names, versions, links, citations) —\n  especially memory and instruction files, which compound errors over time.\n\n## Step 1 — Resolve the target\n\nPick the narrowest target the user named, in this precedence order:\n\n1. **Explicit target** — a file, a directory, a PR diff, or pasted AI output\n   the user pointed at. For a PR: `gh pr diff \u003Cn>` (local) or\n   `mcp__github__pull_request_read` `method: get_diff` (remote\u002Fweb — see\n   `memories\u002Fgithub-mcp-tools.md`); audit only the references the diff *introduces*.\n2. **The memory\u002Finstruction corpus** — the ai-config repo's `memories\u002F`,\n   `skills\u002F`, and `CLAUDE.md`. Find the repo root with\n   `${CLAUDE_PLUGIN_ROOT:-$(git -C ~\u002F.claude\u002Fskills\u002Fpurge-hallucinations rev-parse --show-toplevel 2>\u002Fdev\u002Fnull || pwd)}`.\n3. **The current repo's code & docs** — when the user says \"this repo\" or\n   names no target while inside a project.\n\nIf the target is genuinely ambiguous, ask once; otherwise proceed with the\nnarrowest reasonable reading and state which you chose.\n\n## Step 2 — Extract concrete, checkable references\n\nScan the target for references that have a definite right\u002Fwrong answer. Ignore\nprose, opinions, and design rationale — those aren't hallucinations, just\nclaims. Pull out, with file + line for each:\n\n| Reference type | Examples |\n|---|---|\n| File \u002F path | `src\u002Ffoo.R`, `here(\"data\u002Fx.csv\")`, a relative link in a `.qmd` |\n| Function \u002F object \u002F symbol | `pkg::fn()`, an R object, a shell command, a Make target |\n| Action ref + version | `uses: org\u002Faction@v3`, `@\u003Csha>`, a tag\u002Frelease |\n| Skill name | `skills\u002F\u003Cname>\u002F`, `~\u002F.claude\u002Fskills\u002F\u003Cname>\u002F`, a `\u002Fslash-command` |\n| Memory cross-link | `[[some-memory-name]]` in a memory body |\n| URL \u002F link | `https:\u002F\u002F…`, a docs anchor, a badge target |\n| Citation \u002F package | a CRAN\u002FBioconductor package, a DOI, a `DESCRIPTION` dep |\n| Flag \u002F option \u002F config key | a CLI flag, an env var, a YAML key the schema defines |\n| API \u002F SDK method or model id | an SDK method, a model name, an endpoint |\n\n## Step 3 — Verify each against ground truth\n\nUse the cheapest check that *proves existence or absence*. Match the tool to\nthe reference type:\n\n- **File \u002F path** — `test -e \u003Cpath>`, `git ls-files -- \u003Cpath>`, `ls`.\n- **Function \u002F object** — grep for the *definition*, not just a mention\n  (`grep -rn \"fn \u003C- function\" R\u002F`); for an exported R fn check `NAMESPACE` \u002F\n  `Rscript -e 'library(pkg); exists(\"fn\")'`. Shell command — `command -v`.\n- **Action ref + version** — confirm the tag\u002FSHA exists\n  (`gh api repos\u002F\u003Corg>\u002F\u003Crepo>\u002Fgit\u002Frefs\u002Ftags\u002F\u003Ctag>` or `...\u002Fcommits\u002F\u003Csha>`).\n- **Skill name** — `ls skills\u002F\u003Cname>\u002F` in the local ai-config directory. If\n  not found locally, check the session's available-skills list (appears in\n  system reminders) before classifying as ❌; a globally-available system skill\n  with no local directory is ❓ Unverifiable, not ❌ Fabricated.\n- **Memory cross-link** — `[[target]]` links resolve to **skill directories**\n  (`ls skills\u002F\u003Ctarget>\u002F`); if no matching skill, search memory headings\n  (`grep -rn \"^# .*\u003Ctarget>\" memories\u002F`).\n- **URL \u002F link** — `curl -sSI -o \u002Fdev\u002Fnull -w '%{http_code}' \u003Curl>` (local) or\n  `WebFetch` (remote). **A 404\u002F410 is fabricated; a timeout, 403, 429, or\n  DNS failure is *unverifiable*** — distinguish them.\n- **Citation \u002F package** — CRAN: `https:\u002F\u002Fcran.r-project.org\u002Fpackage=\u003Cpkg>`;\n  installed: `Rscript -e 'find.package(\"\u003Cpkg>\")'`; dep: check `DESCRIPTION`;\n  DOI \u002F `.bib`: `python3 scripts\u002Fcheck_doi_bib.py \u003Cpath-to-.bib>` (or `--root \u003Cdir> --cited-only`).\n- **Flag \u002F option \u002F config key** — grep the tool's `--help`, its schema, or\n  its source; for a YAML key, the consuming code\u002Fschema.\n- **API \u002F SDK method or model id** — check the SDK source\u002Fdocs. For\n  Claude\u002FAnthropic specifics (model ids, params), **defer to the `claude-api`\n  skill** rather than guessing.\n\nSort every reference into exactly one bucket:\n\n| Bucket | Meaning | Disposition |\n|---|---|---|\n| ✅ **Resolves** | proven to exist | leave it |\n| ❌ **Fabricated** | proven absent (404, no definition, no such tag\u002Ffile) | propose a fix (Step 4) |\n| ❓ **Unverifiable** | can't be checked here (network, private, ambiguous) | **report, do not edit** |\n\nWhen in doubt, it's ❓ not ❌. False deletions are worse than a flagged\nmaybe.\n\n## Step 4 — Propose a fix for each fabrication (interactive)\n\nDefault action is **propose, then apply on confirmation** — never bulk-edit\nsilently. For each ❌, present: the reference, where it is, the evidence it's\nfake, and a proposed fix — exactly one of:\n\n1. **Correct it** — when a real target is an obvious near-match (typo, moved\n   path, renamed symbol, bumped version). Show the before→after. Prefer this\n   over deletion when a true referent clearly exists (`grep`\u002F`gh` for the\n   nearest real name).\n2. **Remove it** — when the reference is wholly invented and removing it leaves\n   the sentence\u002Fcode correct.\n3. **Rewrite around it** — when removal would leave a dangling clause or broken\n   code; restate without the fake reference.\n\nApply each fix only after the user confirms (batch obvious typo-corrections\ntogether if the user prefers). Re-verify after editing. If the target is a\nmemory or instruction file, the global *\"verify before recommending\"* rule is\nthe reason this matters — a fabricated memory poisons every future session.\n\n## Step 5 — Report\n\nSummarize: N references checked → ✅ resolved, ❌ fixed (list before→after),\n❓ unverifiable (list, with why each couldn't be checked). The ❓ list is a\nfeature, not a failure — it tells the user exactly what still needs a human\neyeball.\n\n## Custom agent for the detect phase\n\nSteps 1--3 (resolve target, extract references, verify against ground truth)\nhave no need for Edit\u002FWrite access. Delegate them to the\n`hallucination-detector` custom agent (`.claude\u002Fagents\u002Fhallucination-detector.md`)\nfor a hard, harness-enforced guarantee against Edit\u002FWrite tool use before the\nreport in Step 4 is reviewed --- tighter than this skill's own\ninstruction-only discipline, though the agent retains `Bash` for read-only\nchecks, so avoiding a write-capable shell command is still instruction-level.\nRun Step 4 (propose\u002Fapply fixes) in the main session afterward.\n\n## Relationship to other skills\n\n- **`record-learnings`, `ums`, `memorize` \u002F `remember`** — these *write* the\n  memory\u002Fskill\u002Finstruction corpus; `purge-hallucinations` *audits* it. A\n  natural hand-off: after a big memory\u002Fskills write, run this to catch\n  fabricated cross-links and stale paths.\n- **`claude-api`** — the source of truth for Claude\u002FAnthropic model ids and\n  params; defer to it instead of guessing when verifying those references.\n- **`simplify` \u002F `tidy`** — clean *structure* (dead code, redundancy); this\n  cleans *truth* (references that don't resolve). Complementary passes.\n- **`reprexes`** — if verifying a code reference needs actually running it,\n  isolate it as a reprex first.\n- **`check-rendered-refs` \u002F `crr`** — the output-side counterpart. This skill\n  audits *source* references; `crr` scans *rendered* HTML for crossrefs\u002Fcitations\n  that broke at render time and leaked into the page as `?@key`. Hand a `?@key`\n  hit here to trace the dangling key back to its source line.\n- **`check-info-quality` \u002F `ciq`** — this skill checks whether a reference\n  **exists**; `ciq`'s check C checks whether an existing, real reference\n  **actually supports** the claim it's attached to. A citation can pass here\n  and still fail `ciq` if it resolves but doesn't say what the text claims.\n\n## Anti-patterns\n\n- ❌ Deleting an **unverifiable** reference as if it were proven fake — the\n  cardinal sin. Network\u002Fpermission failures are ❓, not ❌.\n- ❌ Flagging prose, opinions, or design rationale — only concrete, checkable\n  references are in scope.\n- ❌ Confirming a symbol exists by finding a *mention* of it rather than its\n  *definition*.\n- ❌ Bulk-editing without proposing fixes first (default is interactive).\n- ❌ Treating a network 403\u002F429\u002Ftimeout the same as a 404.\n- ❌ Guessing whether a Claude\u002FAnthropic model id or param is real instead of\n  deferring to `claude-api`.\n",{"name":5,"description":7,"allowed-tools":24,"user-invocable":29},[25,26,27,28],"Bash","Read","Edit","Write",true,"main","0f720481f8ce3b3d9fef485f6e230d8b3035748e","skills\u002Fpurge-hallucinations",1,"https:\u002F\u002Fgithub.com\u002FMorrison-Lab\u002Fai-config","https:\u002F\u002Fmorrison-lab.github.io\u002Fai-config\u002F","https:\u002F\u002Fgithub.com\u002FMorrison-Lab","2026-11-15 04:39:15",{"source":8,"skillName":5,"commands":39,"prompt":49,"zipAvailable":29},[40,43,46],{"tool":41,"command":42},"npx","npx skills add Morrison-Lab\u002Fai-config --skill purge-hallucinations",{"tool":44,"command":45},"bunx","bunx skills add Morrison-Lab\u002Fai-config --skill purge-hallucinations",{"tool":47,"command":48},"pnpm","pnpm dlx skills add Morrison-Lab\u002Fai-config --skill purge-hallucinations","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FMorrison-Lab\u002Fai-config\u002Fblob\u002Fmain\u002Fskills\u002Fpurge-hallucinations\u002FSKILL.md",[],[52,59,66,73,80,87],{"skillNo":53,"name":54,"nameZh":6,"description":55,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":56,"categories":57,"keywords":58,"publishedAt":19},"01M735BVYCGR5VEC65AFDVSNKW","wrap-up","End-of-session verification and UMS.","https:\u002F\u002Fgithub.com\u002FMorrison-Lab\u002Fai-config\u002Fblob\u002Fmain\u002Fskills\u002Fwrap-up\u002FSKILL.md",[],[],{"skillNo":60,"name":61,"nameZh":6,"description":62,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":63,"categories":64,"keywords":65,"publishedAt":19},"01M735BTH469T1CQ8G2Q9GENCR","workaround-watcher","Watch upstream issue for fix.","https:\u002F\u002Fgithub.com\u002FMorrison-Lab\u002Fai-config\u002Fblob\u002Fmain\u002Fskills\u002Fworkaround-watcher\u002FSKILL.md",[],[],{"skillNo":67,"name":68,"nameZh":6,"description":69,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":70,"categories":71,"keywords":72,"publishedAt":19},"01M735BRKZWFANSQ0GM380B4TD","wn","→ whats-next.","https:\u002F\u002Fgithub.com\u002FMorrison-Lab\u002Fai-config\u002Fblob\u002Fmain\u002Fskills\u002Fwn\u002FSKILL.md",[],[],{"skillNo":74,"name":75,"nameZh":6,"description":76,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":77,"categories":78,"keywords":79,"publishedAt":19},"01M735BPE9V830B3B6NGXQ56J6","whats-next","Recommend concrete next step or action.","https:\u002F\u002Fgithub.com\u002FMorrison-Lab\u002Fai-config\u002Fblob\u002Fmain\u002Fskills\u002Fwhats-next\u002FSKILL.md",[],[],{"skillNo":81,"name":82,"nameZh":6,"description":83,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":84,"categories":85,"keywords":86,"publishedAt":19},"01M735BMHQ60EAD5B7QAWXC5A0","wait-for-results","Wait for pipeline\u002FCI results.","https:\u002F\u002Fgithub.com\u002FMorrison-Lab\u002Fai-config\u002Fblob\u002Fmain\u002Fskills\u002Fwait-for-results\u002FSKILL.md",[],[],{"skillNo":88,"name":89,"nameZh":6,"description":90,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":91,"categories":92,"keywords":93,"publishedAt":19},"01M735BKGZHPN1PEMJ7YBD5VJ8","wai","→ send-notes.","https:\u002F\u002Fgithub.com\u002FMorrison-Lab\u002Fai-config\u002Fblob\u002Fmain\u002Fskills\u002Fwai\u002FSKILL.md",[],[]]