agentleFS
mmaripanguecmmaripanguec/dev-sdd-kit/.claude/skills/repo-map/SKILL.md

repo-map

Skill2 starsChanged 4 months agoMITView on GitHub

Generates or updates a repository's CONTEXT PACK (loadable skill .claude/skills/<prefijo>-<repo>) — architecture, core mechanisms, dependencies and pitfalls with file:line evidence — and seeds its assertions. Use when asked to "map a repo", "generate the repo's context", when /spec-create detects a repo without a pack or with a stale pack, or after /repo-add.

13 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

4 rules. 0 are common, 2 are distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 12 edited versions we read keep these.

  • 2

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Run scripts/freshness.sh check <pack> — it must come out current.

    line 44
  2. Distinctive:

    If while verifying you detect an INCONSISTENCY (the code contradicts the registry, an ADR or an existing pack): STOP AND ASK the human.

    line 45
  3. Distinctive:

    Do not normalize it silently.

    line 47
  4. Very specific:

    Update the mapa-sistemas index if the context state changed (or request /system-map if the system pack does not exist).

    line 50

“Common” is measured: kept by at least 6 in 10 of the 12 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Generate the context pack for repo $ARGUMENTS as a loadable skill: .claude/skills/$(pack_name_for_repo <repo>)/SKILL.md (name = <pack_prefix>-<repo>, or the registry's pack field). The goal: any agent that receives a requirement about this application loads the pack and works with the correct mental model without rereading the repo.

1. Inputs (from cheapest to most expensive)

  1. Registry (repos.yaml): role, domain, vcs; current seal with . scripts/repo-lib.sh && stamp_of_repo <repo>.
  2. As-is facts: knowledge/as-is/<repo>/{modules,api-surface}.md (regenerate if the seal does not match).
  3. codebase-memory MCP graph if indexed (get_architecture, search_graph, trace_path). If it does not respond, continue without it and declare it in "What I DON'T know".
  4. Targeted reading of the code: entrypoint, wiring, routes, data layer, external clients. Read what is needed to back each claim.

2. Write the pack

  • Template: templates/pack-repo.md (replace ALL {{...}} and <...> placeholders; frontmatter with generado_desde: <repo>: <sello> and verificado: with today's date).
  • The frontmatter DESCRIPTION is the load trigger: name it with the application, its topics and typical questions.
  • Hard rules: without file:line evidence a claim does not enter the body (it goes to "What I DON'T know"); the "Pitfalls" and "What I DON'T know" sections are never left empty (if there are no pitfalls yet, say which mistake would be easy to make); budget ~150 lines — detail overflows into references/ inside the pack's folder.
  • Secrets: FORBIDDEN to copy tokens, URLs with credentials or customer data into the pack.

3. Assertions and verification

  • Seed/update scripts/assertions.d/<sistema>.sh with 3+ assertions from the pack's most important claims (format per that folder's README) and run scripts/assertions.sh <sistema> — it must end up green.
  • Run scripts/freshness.sh check <pack> — it must come out current.
  • If while verifying you detect an INCONSISTENCY (the code contradicts the registry, an ADR or an existing pack): STOP AND ASK the human. Do not normalize it silently.

4. Closure

  • Update the mapa-sistemas index if the context state changed (or request /system-map if the system pack does not exist).
  • Commit: docs(packs): pack <pack> @<sello>.
  • Final summary: mental model in 3 lines + pitfalls + what was left out.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.