[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6CCXC4B64ACJWTQ033ECE4Y":3,"skill-same:01M6CCXC4B64ACJWTQ033ECE4Y":89},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":17,"publishedAt":38,"linkOnly":39,"markdownContent":40,"markdownBody":41,"frontMatter":42,"refName":43,"commitSha":44,"skillDirectory":45,"forksCount":46,"repositoryHtmlUrl":47,"repositoryHomepageUrl":48,"ownerHtmlUrl":49,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":50,"install":51,"resources":64},"01M6CCXC4B64ACJWTQ033ECE4Y","severed-wire-audit",null,"Find and resolve \"severed wires\" in a codebase — a producer and a consumer both exist and are fully implemented, but the connecting registration \u002F dispatch arm \u002F event subscription \u002F call site between them is missing, so a feature looks broken though both ends are done. dead-code lints CANNOT catch this: the producer isn't dead (it has tests) and the consumer connects to a stub or reads an inert config. Use when a feature \"looks unwired \u002F 没接上 \u002F 两端都在但连线断 \u002F 明明写了却调不到\"; after a refactor or middleware removal, to verify producers still reach consumers; when asked to \"审接缝 \u002F 找断线 \u002F seam audit \u002F wire audit \u002F connection audit\"; or when a method returns NOT_FOUND, a tool the model can't call, an event that never fires, or a config field nothing reads. Triages each finding into CONNECT (fix the wire) \u002F CUT (delete the dead abstraction) \u002F DECIDE (ask the human), enforces read-before-write so dead scaffolding is deleted not reconnected, and installs a grep-diff guard against regression.","rootazero\u002FAleph","rootazero","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F124485722?v=4","01M6CCWSEZCM9XS2W4R9XR5YGH","01M6CCWSK4WJ6D0W8VRJNSZZKA",4,"MIT","https:\u002F\u002Fgithub.com\u002Frootazero\u002FAleph\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fsevered-wire-audit\u002FSKILL.md",[],[18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37],"severed wire audit","audit","decide","dispatch","never","diff","use","fix","read","ask","broken","connect","find","regression","model","tool","guard","verify","code","delete","2026-11-10 10:12:30",false,"---\nname: severed-wire-audit\ndescription: >-\n  Find and resolve \"severed wires\" in a codebase — a producer and a consumer both\n  exist and are fully implemented, but the connecting registration \u002F dispatch arm \u002F\n  event subscription \u002F call site between them is missing, so a feature looks broken\n  though both ends are done. dead-code lints CANNOT catch this: the producer isn't\n  dead (it has tests) and the consumer connects to a stub or reads an inert config.\n  Use when a feature \"looks unwired \u002F 没接上 \u002F 两端都在但连线断 \u002F 明明写了却调不到\"; after a\n  refactor or middleware removal, to verify producers still reach consumers; when\n  asked to \"审接缝 \u002F 找断线 \u002F seam audit \u002F wire audit \u002F connection audit\"; or when a\n  method returns NOT_FOUND, a tool the model can't call, an event that never fires,\n  or a config field nothing reads. Triages each finding into CONNECT (fix the wire) \u002F\n  CUT (delete the dead abstraction) \u002F DECIDE (ask the human), enforces\n  read-before-write so dead scaffolding is deleted not reconnected, and installs a\n  grep-diff guard against regression.\n---\n\n# Severed Wire Audit\n\n## The problem this solves (病灶)\n\nAcross a codebase, features repeatedly **look missing** when the truth is narrower:\na **producer** exists, a **consumer** exists, and only the **wire between them is cut**.\n两端都在，连线断。\n\nThe trap: `dead_code` \u002F unused-import lints only catch **form 1** (a producer with no\nconsumer). They are blind to the four forms that matter more — a wire whose far end is a\n**stub**, a config nobody **reads**, a client call with no **handler**, a name\u002Fpath that\n**drifted**. So the compiler stays green while the feature is dark.\n\nThe cure is to **audit seams, not modules**, then triage each severed wire — and the\nsingle most expensive mistake is assuming a severed wire should be *reconnected*. Very\noften it should be **cut**.\n\n## The five-phase workflow\n\nRun these in order. Do not skip phase 3's read-first rule — it is where most of the value (and most of the avoided damage) lives.\n\n### 1. Scan seams (multi-lens, parallel, read-only)\n\nA seam is a join point where one side's output must reach the other side's input. Sweep\neach seam type as an independent grep-diff of **PRODUCED\u002FDEFINED symbols** minus\n**CONSUMED\u002FDISPATCHED symbols**.\n\n**Scope first.** Audit one subsystem's seams when the ask is \"审计 X 子系统\"; go whole-repo\nwhen it's \"find all severed wires\" or after a broad refactor \u002F middleware removal.\n\n**Fan out one read-only agent per seam type.** They don't interfere, and one grep angle\nalways misses what another catches (the 2026-07-15 audit ran seven in parallel). Give each a\nprompt shaped like: *\"Under `\u003Cpath>`, list every `\u003CDEFINED pattern>` and every `\u003CCONSUMED\npattern>`; return the set difference with a file:line for both ends. Read-only — do not fix\nor edit anything.\"* Collect every agent's diff into the phase-2 list.\n\nThe seam types and the \"6 forms of severed wire\" (with concrete cases) are in\n**[references\u002Fseam-catalog.md](references\u002Fseam-catalog.md)** — read it before scanning so\nyou know what each lens targets.\n\n### 2. Enumerate candidates into one list\n\nCollect every `DEFINED − CONSUMED` mismatch into a single flat list with a file:line\nanchor for **both** ends (or \"no consumer found\"). Do not fix anything yet. The list is\nalmost always longer than the real defect count — that's expected, phase 3 prunes it.\n\n### 3. Triage each candidate — READ-FIRST, then CONNECT \u002F CUT \u002F DECIDE\n\nThis is the heart of the skill. **Both** the CONNECT and the CUT buckets systematically\n**over-count**, in opposite directions, and the only defense is to read the actual\nconsumers\u002Fcallers before touching anything.\n\n- **The read-before-write rule (non-negotiable):** the first action on any candidate is\n  to grep for a **live caller** on the *consumer* side (`Api::x(`, `rpc_call(\"x\")`,\n  `.subscribe(`, an actual field read). A \"missing handler\" is very often a **dead client\n  wrapper** with zero callers → that's a **CUT**, not a CONNECT. Reconnecting dead\n  scaffolding is worse than useless: it revives an API that then needs perpetual syncing.\n\n- **The painless-wire heuristic:** a wire that has been severed a long time with **zero\n  observed production pain** — 一根断了却没人喊疼的线 — is a candidate to **CUT**, not\n  reconnect. Especially if an error path or a newer mechanism already covers the same\n  outcome.\n\n- **DECIDE** is for genuine feature calls where reconnecting means real new coupling, or\n  where \"delete vs revive\" is a product judgment. Present the trade-off and ask; don't\n  silently pick.\n\nThe full decision tree, the \"systematic over-count\" meta-finding, and the real\ncase studies (5 candidates that read as \"broken wires\" but were dead scaffolding) are in\n**[references\u002Ftriage-playbook.md](references\u002Ftriage-playbook.md)**.\n\n### 4. Fix — surgically, and grep every symbol before deleting a file\n\n- **CONNECT:** add the one missing wire (register the tool, add the dispatch arm, wire the\n  subscription, point the client at the real method). Nothing else.\n- **CUT:** delete the dead producer *and* its now-orphaned consumer stub, tests, docs, and\n  imports — but **grep every exported symbol of a file before deleting the file**, not just\n  the headline type. A file named for a dead tool can still export an enum a live config\n  field consumes (the `CleanupPolicy` E0432 lesson: deleting the tool broke the config\n  schema). If a to-be-deleted symbol has a live consumer, **relocate** it to that consumer\n  rather than deleting.\n- Keep diffs minimal and match surrounding style (see the user's global rules).\n- **Verify with a pass that compiles TEST code**, not just a fast type-check — a CUT can\n  break a `#[cfg(test)]` consumer that `cargo check --lib` never compiles (use\n  `cargo test --no-run`, or the equivalent in your language). Then re-run the phase-5 guard:\n  green means nothing severed beyond the baseline.\n\n### 5. Guard — turn the runtime convention into a compile-time or CI check\n\nA severed wire recurs because the same fact is copied into ≥3 unsynchronized lists with\n**no single source of truth**. \"No omission\" is only real when the convention is\n*enforced*, not documented:\n\n- **Best — compile-time:** one enum\u002Ftype as the single true source, with the other lists\n  *derived* from it, so an unrouted method fails to compile. (In Aleph: RPC method single\n  enum; a tool-registration completeness test; a config-reader completeness test.)\n- **Good — CI grep-diff guard:** a script computing `DEFINED − CONSUMED` against a\n  `KNOWN_SEVERED` baseline, wired into the test target. **Fix one → remove one from the\n  baseline.** A new severed wire then turns the check red.\n\nThe grep-diff guard is a reusable, parametrized script:\n**[scripts\u002Fwiring_audit.py](scripts\u002Fwiring_audit.py)** — point it at your DEFINED pattern +\nglob and your CONSUMED pattern + files. The design rationale, baseline discipline, and the\n\"a guard is a **detector, not a judge**\" rule (it flags inert wires; CONNECT-vs-CUT still\nneeds a human) are in **[references\u002Fguard-scripts.md](references\u002Fguard-scripts.md)**.\n\n## The four non-negotiable rules (checklist)\n\nCreate one todo per rule when running an audit:\n\n1. **Read before write.** Grep for a live *caller\u002Fconsumer* before deciding CONNECT vs CUT.\n   No live caller ⇒ default CUT.\n2. **Grep every exported symbol before deleting a file** — not just the headline type.\n   Relocate symbols that have a live consumer.\n3. **A painless severed wire is a CUT candidate.** Absence of production pain is evidence,\n   not a coincidence to preserve.\n4. **A guard is a detector, not a judge.** Grep-diff catches what LLM audits miss\n   (in Aleph it caught `memory.store` and `AiRetrievalPolicy` that the semantic sweep\n   dropped) — but whether to CONNECT or CUT each flagged wire is a human\u002Fread decision.\n\n## Why not just trust dead-code lints or an LLM sweep?\n\n- Lints see only form 1. Stubs, inert config, client-ghost, and name-drift stay green.\n- An LLM semantic sweep over-trusts the *handler* side (\"there's a stub → it's a broken\n  wire to fix\") and under-checks the *client* side. It also silently misses entries — the\n  grep-diff guard is what makes the audit exhaustive. Use both: LLM lenses for judgment,\n  grep-diff for completeness.\n","# Severed Wire Audit\n\n## The problem this solves (病灶)\n\nAcross a codebase, features repeatedly **look missing** when the truth is narrower:\na **producer** exists, a **consumer** exists, and only the **wire between them is cut**.\n两端都在，连线断。\n\nThe trap: `dead_code` \u002F unused-import lints only catch **form 1** (a producer with no\nconsumer). They are blind to the four forms that matter more — a wire whose far end is a\n**stub**, a config nobody **reads**, a client call with no **handler**, a name\u002Fpath that\n**drifted**. So the compiler stays green while the feature is dark.\n\nThe cure is to **audit seams, not modules**, then triage each severed wire — and the\nsingle most expensive mistake is assuming a severed wire should be *reconnected*. Very\noften it should be **cut**.\n\n## The five-phase workflow\n\nRun these in order. Do not skip phase 3's read-first rule — it is where most of the value (and most of the avoided damage) lives.\n\n### 1. Scan seams (multi-lens, parallel, read-only)\n\nA seam is a join point where one side's output must reach the other side's input. Sweep\neach seam type as an independent grep-diff of **PRODUCED\u002FDEFINED symbols** minus\n**CONSUMED\u002FDISPATCHED symbols**.\n\n**Scope first.** Audit one subsystem's seams when the ask is \"审计 X 子系统\"; go whole-repo\nwhen it's \"find all severed wires\" or after a broad refactor \u002F middleware removal.\n\n**Fan out one read-only agent per seam type.** They don't interfere, and one grep angle\nalways misses what another catches (the 2026-07-15 audit ran seven in parallel). Give each a\nprompt shaped like: *\"Under `\u003Cpath>`, list every `\u003CDEFINED pattern>` and every `\u003CCONSUMED\npattern>`; return the set difference with a file:line for both ends. Read-only — do not fix\nor edit anything.\"* Collect every agent's diff into the phase-2 list.\n\nThe seam types and the \"6 forms of severed wire\" (with concrete cases) are in\n**[references\u002Fseam-catalog.md](references\u002Fseam-catalog.md)** — read it before scanning so\nyou know what each lens targets.\n\n### 2. Enumerate candidates into one list\n\nCollect every `DEFINED − CONSUMED` mismatch into a single flat list with a file:line\nanchor for **both** ends (or \"no consumer found\"). Do not fix anything yet. The list is\nalmost always longer than the real defect count — that's expected, phase 3 prunes it.\n\n### 3. Triage each candidate — READ-FIRST, then CONNECT \u002F CUT \u002F DECIDE\n\nThis is the heart of the skill. **Both** the CONNECT and the CUT buckets systematically\n**over-count**, in opposite directions, and the only defense is to read the actual\nconsumers\u002Fcallers before touching anything.\n\n- **The read-before-write rule (non-negotiable):** the first action on any candidate is\n  to grep for a **live caller** on the *consumer* side (`Api::x(`, `rpc_call(\"x\")`,\n  `.subscribe(`, an actual field read). A \"missing handler\" is very often a **dead client\n  wrapper** with zero callers → that's a **CUT**, not a CONNECT. Reconnecting dead\n  scaffolding is worse than useless: it revives an API that then needs perpetual syncing.\n\n- **The painless-wire heuristic:** a wire that has been severed a long time with **zero\n  observed production pain** — 一根断了却没人喊疼的线 — is a candidate to **CUT**, not\n  reconnect. Especially if an error path or a newer mechanism already covers the same\n  outcome.\n\n- **DECIDE** is for genuine feature calls where reconnecting means real new coupling, or\n  where \"delete vs revive\" is a product judgment. Present the trade-off and ask; don't\n  silently pick.\n\nThe full decision tree, the \"systematic over-count\" meta-finding, and the real\ncase studies (5 candidates that read as \"broken wires\" but were dead scaffolding) are in\n**[references\u002Ftriage-playbook.md](references\u002Ftriage-playbook.md)**.\n\n### 4. Fix — surgically, and grep every symbol before deleting a file\n\n- **CONNECT:** add the one missing wire (register the tool, add the dispatch arm, wire the\n  subscription, point the client at the real method). Nothing else.\n- **CUT:** delete the dead producer *and* its now-orphaned consumer stub, tests, docs, and\n  imports — but **grep every exported symbol of a file before deleting the file**, not just\n  the headline type. A file named for a dead tool can still export an enum a live config\n  field consumes (the `CleanupPolicy` E0432 lesson: deleting the tool broke the config\n  schema). If a to-be-deleted symbol has a live consumer, **relocate** it to that consumer\n  rather than deleting.\n- Keep diffs minimal and match surrounding style (see the user's global rules).\n- **Verify with a pass that compiles TEST code**, not just a fast type-check — a CUT can\n  break a `#[cfg(test)]` consumer that `cargo check --lib` never compiles (use\n  `cargo test --no-run`, or the equivalent in your language). Then re-run the phase-5 guard:\n  green means nothing severed beyond the baseline.\n\n### 5. Guard — turn the runtime convention into a compile-time or CI check\n\nA severed wire recurs because the same fact is copied into ≥3 unsynchronized lists with\n**no single source of truth**. \"No omission\" is only real when the convention is\n*enforced*, not documented:\n\n- **Best — compile-time:** one enum\u002Ftype as the single true source, with the other lists\n  *derived* from it, so an unrouted method fails to compile. (In Aleph: RPC method single\n  enum; a tool-registration completeness test; a config-reader completeness test.)\n- **Good — CI grep-diff guard:** a script computing `DEFINED − CONSUMED` against a\n  `KNOWN_SEVERED` baseline, wired into the test target. **Fix one → remove one from the\n  baseline.** A new severed wire then turns the check red.\n\nThe grep-diff guard is a reusable, parametrized script:\n**[scripts\u002Fwiring_audit.py](scripts\u002Fwiring_audit.py)** — point it at your DEFINED pattern +\nglob and your CONSUMED pattern + files. The design rationale, baseline discipline, and the\n\"a guard is a **detector, not a judge**\" rule (it flags inert wires; CONNECT-vs-CUT still\nneeds a human) are in **[references\u002Fguard-scripts.md](references\u002Fguard-scripts.md)**.\n\n## The four non-negotiable rules (checklist)\n\nCreate one todo per rule when running an audit:\n\n1. **Read before write.** Grep for a live *caller\u002Fconsumer* before deciding CONNECT vs CUT.\n   No live caller ⇒ default CUT.\n2. **Grep every exported symbol before deleting a file** — not just the headline type.\n   Relocate symbols that have a live consumer.\n3. **A painless severed wire is a CUT candidate.** Absence of production pain is evidence,\n   not a coincidence to preserve.\n4. **A guard is a detector, not a judge.** Grep-diff catches what LLM audits miss\n   (in Aleph it caught `memory.store` and `AiRetrievalPolicy` that the semantic sweep\n   dropped) — but whether to CONNECT or CUT each flagged wire is a human\u002Fread decision.\n\n## Why not just trust dead-code lints or an LLM sweep?\n\n- Lints see only form 1. Stubs, inert config, client-ghost, and name-drift stay green.\n- An LLM semantic sweep over-trusts the *handler* side (\"there's a stub → it's a broken\n  wire to fix\") and under-checks the *client* side. It also silently misses entries — the\n  grep-diff guard is what makes the audit exhaustive. Use both: LLM lenses for judgment,\n  grep-diff for completeness.\n",{"name":5,"description":7},"main","ad8cf9736a137eb206d78e1db13ac9d452fb27ac",".claude\u002Fskills\u002Fsevered-wire-audit",1,"https:\u002F\u002Fgithub.com\u002Frootazero\u002FAleph","https:\u002F\u002Fwww.heyaleph.com","https:\u002F\u002Fgithub.com\u002Frootazero","2026-11-12 05:58:43",{"source":8,"skillName":5,"commands":52,"prompt":62,"zipAvailable":63},[53,56,59],{"tool":54,"command":55},"npx","npx skills add rootazero\u002FAleph --skill severed-wire-audit",{"tool":57,"command":58},"bunx","bunx skills add rootazero\u002FAleph --skill severed-wire-audit",{"tool":60,"command":61},"pnpm","pnpm dlx skills add rootazero\u002FAleph --skill severed-wire-audit","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Frootazero\u002FAleph\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fsevered-wire-audit\u002FSKILL.md",true,[65,71,75,79,83],{"path":66,"resourceType":67,"resourceTypeDesc":68,"sizeBytes":69,"sourceHtmlUrl":70},"index.md","REFERENCE","参考文档",483,"https:\u002F\u002Fgithub.com\u002Fxberg-io\u002Fliter-llm\u002Fblob\u002Fmain\u002Fplugin\u002Fskills\u002Fliter-llm-using-the-mcp-server\u002Findex.md",{"path":72,"resourceType":67,"resourceTypeDesc":68,"sizeBytes":73,"sourceHtmlUrl":74},"references\u002Fguard-scripts.md",5212,"https:\u002F\u002Fgithub.com\u002Frootazero\u002FAleph\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fsevered-wire-audit\u002Freferences\u002Fguard-scripts.md",{"path":76,"resourceType":67,"resourceTypeDesc":68,"sizeBytes":77,"sourceHtmlUrl":78},"references\u002Fseam-catalog.md",5806,"https:\u002F\u002Fgithub.com\u002Frootazero\u002FAleph\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fsevered-wire-audit\u002Freferences\u002Fseam-catalog.md",{"path":80,"resourceType":67,"resourceTypeDesc":68,"sizeBytes":81,"sourceHtmlUrl":82},"references\u002Ftriage-playbook.md",7126,"https:\u002F\u002Fgithub.com\u002Frootazero\u002FAleph\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fsevered-wire-audit\u002Freferences\u002Ftriage-playbook.md",{"path":84,"resourceType":85,"resourceTypeDesc":86,"sizeBytes":87,"sourceHtmlUrl":88},"scripts\u002Fwiring_audit.py","SCRIPT","脚本",8305,"https:\u002F\u002Fgithub.com\u002Frootazero\u002FAleph\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fsevered-wire-audit\u002Fscripts\u002Fwiring_audit.py",[90,98,124,135,144,149],{"skillNo":91,"name":92,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":93,"categories":94,"keywords":95,"publishedAt":38},"01M6CCYRW46NCCDBJVXWDRMCFY","invalid-yaml","https:\u002F\u002Fgithub.com\u002Frootazero\u002FAleph\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fmarkdown_skills\u002Finvalid-yaml\u002FSKILL.md",[],[96,97],"invalid yaml","yaml",{"skillNo":99,"name":100,"nameZh":6,"description":101,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":102,"categories":103,"keywords":109,"publishedAt":38},"01M6CCYNTM3981CQE5P8RXB6Y1","gh-pr-docker","GitHub PR operations (Aleph enhanced with Docker)","https:\u002F\u002Fgithub.com\u002Frootazero\u002FAleph\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fmarkdown_skills\u002Fgh-pr-docker\u002FSKILL.md",[104],{"categoryNo":105,"categoryName":106,"taxonomyType":107,"taxonomyTypeDesc":108},"01JBSEEDCAT000000000000011","运维与部署","SCENARIO","使用场景",[110,111,112,113,114,115,116,117,118,119,120,121,122,123],"gh pr docker","gh","gh pr","pr","docker","operations","view","list","write","format","repository","create","github","cli",{"skillNo":125,"name":126,"nameZh":6,"description":127,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":128,"categories":129,"keywords":130,"publishedAt":38},"01M6CCYME2FV26WMXFJHWDC96R","echo-basic","Basic echo command (frontmatter smoke fixture)","https:\u002F\u002Fgithub.com\u002Frootazero\u002FAleph\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fmarkdown_skills\u002Fecho-basic\u002FSKILL.md",[],[131,132,133,134],"echo basic","Echo","smoke","command",{"skillNo":136,"name":137,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":138,"categories":139,"keywords":140,"publishedAt":38},"01M6CCXQT84VNRDQE74JJXSYR8","review-results","https:\u002F\u002Fgithub.com\u002Frootazero\u002FAleph\u002Fblob\u002Fmain\u002Freview-results\u002Fskill.md",[],[141,142,143],"review results","results","review",{"skillNo":145,"name":137,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":146,"categories":147,"keywords":148,"publishedAt":38},"01M6CCXFTHNVKMMCP7NTMN5HBM","https:\u002F\u002Fgithub.com\u002Frootazero\u002FAleph\u002Fblob\u002Fmain\u002Fdocs\u002Fengineering-reports\u002Freview-results\u002Fskill.md",[],[141,143,142],{"skillNo":150,"name":151,"nameZh":6,"description":152,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":153,"categories":154,"keywords":155,"publishedAt":38},"01M6CCX91ZGRZQPDENFQY66PX6","rust-occams-razor","Use when refactoring, simplifying, reviewing, or decomposing an existing Rust codebase while preserving observable behavior, public contracts, concurrency semantics, and performance characteristics.","https:\u002F\u002Fgithub.com\u002Frootazero\u002FAleph\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Frust-occams-razor\u002FSKILL.md",[],[156,157,158,159,160,161,24,162],"rust","rust occams razor","behavior","concurrency","refactoring","public","performance"]