repo-audit
Skill4 starsChanged 15 days agoMITView on GitHub
Autonomous repo-wide audit (MANIFEST queue). Loops all pending units in one invocation without human checkpoints between units; generates HANDOFF when done. Coherence, broken symlinks, secrets, vault PARA, scripts, doc condensation. Writes to 06_Metadata/Reference/Repo-Audit/. Readonly except audit artifacts. Use with /repo-audit or repo-auditor subagent.
25 other projects use a version of this.
Adopted by
Most recently taken up by
- Edzard9/stashtheir own version, first seen Nov 16, 2026
- albkhachikyan/claude-code-repo-audittheir own version, first seen Nov 15, 2026
- iuliandita/skillstheir own version, first seen Nov 15, 2026
- idaibin/skillstheir own version, first seen Nov 8, 2026
- ekajaya740/skillstheir own version, first seen Nov 1, 2026
- 0xUrsanomics/utopia-ostheir own version, first seen Nov 1, 2026
Used with
Projects with it also use code-review.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
GoogleSkillcode-review
Adversarial systems, concurrency, memory-model, and architecture code review protocol. Use this skill whenever asked to review code, PRs, diffs, or branches in Filament.
Also in 3 projects that use this
Other versions25 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 25
- zebbern/claude-code-guide★ 4.7ksame name
- Kanevry/session-orchestrator★ 53same name
- iuliandita/skills★ 12same name
- Megaprompting/torque-loop★ 10same name
- sunitghub/canon-skills★ 9same name
- idaibin/skills★ 6same name
- my-chiefmind/ai-maestro★ 6same name
- nikkogibler/cabinet★ 4same name
- conorbronsdon/repo-audit★ 3same name
- 0xUrsanomics/utopia-os★ 2same name
15 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository131 other files it gives its agents.
- agent-contractSkill
- brand-chrome-loginSkill
- contexto-aristasSkill
- crear-bot-nicoSkill
- daily-standupSkill
- design-a-grok-botSkill
- drift-hunt-rosterSkill
- explain-grokbotSkill
- exportar-botSkill
- govern-agentic-osSkill
- grok-bot-project-opsSkill
- hallazgoSkill
- handoff-local-hermesSkill
- identity-pipelineSkill
- loop-corpSkill
- no-devSkill
- orca-handoffSkill
- panorama-de-flujoSkill
- reporte-gmSkill
- routine-healthcheckSkill
- sprint-parallelSkill
- transcript-healthcheckSkill
- getting-started-brandySkill
- getting-started-calendario-contenidoSkill
- getting-started-cazador-de-canalesSkill
- getting-started-cazador-de-vuelosSkill
- getting-started-cazador-gangasSkill
- getting-started-cazador-inmobiliarioSkill
- getting-started-cazador-programasSkill
- getting-started-cazador-sponsorsSkill
- getting-started-cazador-viviendaSkill
- getting-started-content-drafterSkill
- getting-started-data-scienceSkill
- getting-started-demand-hunterSkill
- getting-started-digital-storesSkill
- getting-started-directory-builderSkill
- getting-started-enrichment-deskSkill
- getting-started-event-hunterSkill
- getting-started-expense-managerSkill
- getting-started-gm-operacionesSkill
- getting-started-inbox-managerSkill
- getting-started-intake-intelSkill
- getting-started-intelligence-gmSkill
- getting-started-join-a-startupSkill
- getting-started-last30daysSkill
- getting-started-life-opsSkill
- getting-started-luma-taggingSkill
- getting-started-market-watcherSkill
- getting-started-meta-vaultSkill
- getting-started-musa-designerSkill
- getting-started-plantillas-notionSkill
- getting-started-product-hunterSkill
- getting-started-publisherSkill
- getting-started-pulso-brandSkill
- getting-started-pulso-diaSkill
- getting-started-renewals-hunterSkill
- getting-started-saves-archivistSkill
- getting-started-scraping-deskSkill
- getting-started-seo-deskSkill
- getting-started-sliderSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
1 rule. 0 are common, 1 is distinctive, and 0 are very specific to this project.
0
Common
Most of the 21 edited versions we read keep these.
1
Distinctive
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 39
No reimplementes esas skills.
“Common” is measured: kept by at least 6 in 10 of the 21 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Repo Audit
Orquestador de auditoría readonly del vault Agentic OS.
Default: autónomo. Una invocación de /repo-audit (sin args) recorre todas las unidades pending del MANIFEST, una tras otra, sin pedir confirmación al usuario entre unidades. Al terminar la cola, genera HANDOFF.md automáticamente.
Objetivos (todos obligatorios por unidad)
- Coherencia — doc dice X, código/config hace Y.
- Rotos — symlinks, paths muertos, commands sin SKILL, skills sin doc.
- Secrets — riesgo de exposición,
.gitignore, duplicación de credenciales. - Vault — PARA+, frontmatter, wikilinks (cuando aplique a la unidad).
- Scripts —
.scripts/,.config/(calidad/arquitectura, no estilo). - Condensación — qué contenido sobra, está duplicado o puede vivir en menos archivos canónicos. Actualizar
CONDENSATION-MAP.mdcuando encuentres overlap.
Skills satélite (cargar según unidad)
| Unidad tipo | Leer primero |
.claude/, 06_Metadata/Skills|Commands|MCP | Patrón en .claude/agents/sync-auditor.md |
00_–06_, wikilinks | .claude/agents/vault-explorer.md |
.scripts/, .config/, trigger/ | .agents/skills/codebase-design/SKILL.md (secciones relevantes) |
| Doc vs realidad | .agents/skills/grill-with-docs/SKILL.md |
.agents/, skills-lock.json | .agents/skills/supply-chain-audit/SKILL.md (symlink/install lens) |
| Handoff final | .agents/skills/handoff/SKILL.md + references/handoff-template.md + references/executor-prompt-template.md |
No reimplementes esas skills; delega criterios y cita paths concretos.
Artefactos (única zona writable)
06_Metadata/Reference/Repo-Audit/
├── MANIFEST.md # cola de unidades
├── PROGRESS.md # log de sesiones
├── CONDENSATION-MAP.md # mapa vivo de duplicados / merges propuestos
├── sessions/NNN-<slug>.md
└── HANDOFF.md # auto al vaciar pending (modo all)Comando /repo-audit
Interpreta $ARGUMENTS:
| Args | Acción |
(vacío) o all | Modo autónomo (default): auditar todas las unidades pending en secuencia; al vaciar cola → HANDOFF.md. No preguntar al usuario entre unidades. |
next | Solo la siguiente unidad pending (escape hatch / debug) |
status | Resumir progreso + unidades pending (no auditar) |
handoff | Generar o regenerar HANDOFF.md (solo si MANIFEST 100% done, salvo --force implícito por usuario) |
<unit-id> ej. 001 | Auditar solo esa unidad |
Modo autónomo (all) — protocolo obligatorio
- Cargar
.agents/skills/repo-audit/references/condensation-rubric.md. - Ejecutar
bash .agents/skills/repo-audit/scripts/list-pending.sh→ cola ordenada. - Para cada línea
id|slug|path|status:- Ejecutar protocolo de unidad (abajo).
- Actualizar MANIFEST →
done, PROGRESS, CONDENSATION-MAP si aplica. - Continuar inmediatamente con la siguiente unidad — sin resumir al usuario ni pedir "¿sigo?".
- Si la cola queda vacía → generar
HANDOFF.md(template enreferences/handoff-template.md). - Un solo mensaje final al usuario: unidades completadas en esta corrida, conteo P0/P1, path de HANDOFF o checkpoint.
Checkpoint por contexto: si el contexto se agota mid-run, termina la unidad en curso, marca en PROGRESS checkpoint — resume with /repo-audit, y para. La próxima invocación vacía retoma desde list-pending.sh.
Subagentes (Cursor Task): en modo all, puedes lanzar repo-auditor por unidad en la misma sesión padre solo si el host lo permite; preferir secuencial en el mismo agente para no perder CONDENSATION-MAP coherente. Nunca delegar una sola unidad y parar si quedan pending sin avisar en PROGRESS.
Protocolo por unidad
- Resolver unidad:
bash .agents/skills/repo-audit/scripts/next-unit.sh [unit-id](o la fila actual del loop). - Solo leer paths de esa unidad (+ referencias cruzadas mínimas).
- Escribir
sessions/NNN-<slug>.mdusandoreferences/session-report-template.md. - Actualizar fila en
MANIFEST.md→done+ fecha. - Añadir entrada corta en
PROGRESS.md. - Si hay overlap de docs, añadir/actualizar filas en
CONDENSATION-MAP.md. - Modo
all: volver al paso 1 con la siguiente pending — no esperar al usuario.
Unidad 001-root-files (obligatorio especial)
Auditar todos los archivos sueltos en raíz del repo. Para entry points, producir tabla de condensación:
| Archivo | Rol declarado | Overlap con | ¿Canónico? | Acción propuesta |
Candidatos típicos a revisar juntos: CLAUDE.md, AGENTS.md, INDEX.md, README.md, SETUP_COMPLETE.md, AGENTIC_OS_DASHBOARD.md, IDEA.md, .cursorrules.
Regla de condensación: un solo Memory Layer canónico (CLAUDE.md + .claude/docs/*.md). Todo lo demás debe ser puntero corto o eliminable sin pérdida.
Profundidad (pre-order)
- Tras completar una carpeta, si tiene subcarpetas no listadas en MANIFEST, añadir filas
pendingantes de seguir (no saltar profundidad). - Ignorar:
.git/,node_modules/, binarios,.DS_Store. - Skim permitido:
.obsidian/plugins/**/main.js(vendor) — solo coherencia de config, no review de código.
Severidad de hallazgos
- P0 blocker — rompe workflow (symlink roto, path muerto en command activo)
- P1 fix — desalineación doc/código, duplicación costosa
- P2 nit — estilo, typo, doc stale menor
Handoff final
Cuando MANIFEST = 100% done (automático al final de modo all, o manual con handoff):
- Leer todas las
sessions/*.md+CONDENSATION-MAP.md(+DEEP-AUDIT-*.mdsi existen). - Escribir
HANDOFF.mdconreferences/handoff-template.md. - Agrupar fixes por área; sección dedicada "Documentation Condensation Plan" con merges propuestos ordenados por impacto/tokens.
- Listar skills sugeridas para el agente ejecutor (implement + handoff + sync-auditor + intent-layer para AGENTS.md hierarchy).
- Obligatorio — §9 al final: cerrar HANDOFF con
## 9. Executor handoff→### 9.1 Prompt copy-paste. Rellenar el bloquetextusandoreferences/executor-prompt-template.md({{P0_LIST}},{{P1_LIST}},{{CONDENSATION_TOP3}}con datos reales de este handoff). Nada después de §9.1 — es la última sección del archivo. - En el mensaje final al usuario, indicar: "Prompt ejecutor en HANDOFF §9.1 — copiar a un chat nuevo para implementar."
No es tu trabajo
- Aplicar fixes en el repo (otro agente lo hace).
- Borrar archivos "duplicados" sin fila en CONDENSATION-MAP y aprobación humana implícita vía handoff.
- Regenerar
06_Metadatavíasync-docs.sh(solo recomendar en handoff). - Pedir confirmación humana entre unidades en modo default (
all).
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

