agentsclimarketplace

Codemap

Skill szobonyaerik/agentic-avengers/skills/codemap

Generate and read the precomputed codebase map that grounds the planning agents. Use at setup time and whenever an agent needs the real module structure, dependency flow, and entry points of the project — especially before architecture, planning, or provisioning. The map is produced deterministically with tree-sitter and works without any LLM. Always consult the codemap before assuming a file path or module boundary.From its SKILL.md

Install
npx -y skills add szobonyaerik/agentic-avengers --skill codemap

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

3.4 KB, as published. Nobody here has run it

codemap

The planning agents in this pipeline are stack-agnostic — all of their awareness of this project's structure comes from the codemap. Treat codebase/MOC.md as ground truth for module boundaries, dependency direction, and entry points. Never invent structure; read the map.

Prerequisites (once per environment, per language)

pip install tree-sitter tree-sitter-python      # + tree-sitter-java / tree-sitter-c as needed
pip install rich                                # optional: progress bar + summary

Generate the map

Run from the project root, writing into codebase/ so the agents find it:

python scripts/codemap.py . --lang <python|java|c> --output codebase

Outputs under codebase/: MOC.md (the map the agents read), INDEX.md, one detail doc per source file, and .codemap-manifest.json (cache — do not hand-edit).

  • Re-run after changes; hand-written notes inside each file doc's <!-- notes --> block survive.
  • Full rebuild: add --force — but see the caveat below, it discards cached purposes.
  • Each module's purpose comes from its docstring / KDoc / Javadoc; a file that documents itself gets a real purpose line, one that does not is marked (undocumented).

The optional LLM purpose backfill is currently disabled. Do not pass --provider, --model, --base-url or --api-key — they now exit non-zero. --no-llm is still accepted but is a no-op, and --changed has no effect while purposes are disabled. An existing .codemap-manifest.json is still read, so purposes an earlier model-backed run cached keep rendering as long as the file is unchanged; it is never refreshed or rewritten. This is why --force is a bad default — it throws those cached purposes away.

Per-language status

  • Python — rich (tree-sitter): exports, imports, module-path dependency graph.
  • Java — supported (tree-sitter): FQN declarations + import graph; Gradle module discovery.
  • Kotlin — supported (tree-sitter), if a team uses it.
  • C — supported (tree-sitter, .c/.h): structural extraction via #include/declarations. Note this is text-structural, not compiler-driven, so it's thinner on heavy preprocessor/macro use.
  • C++ — NOT supported. There is no cpp LanguageSpec (.cpp/.hpp, tree_sitter_cpp, namespaces, templates). A C++ team needs that spec added first; even then templates/preprocessor limit fidelity, so a C++ team should lean harder on the human spec review to compensate.

How agents use it

  • solution-architect / implementation-planner / spec-writer read codebase/MOC.md to ground architecture, phasing, and spec paths in reality.
  • agent-factory reads it to detect the stack and whether a UI surface exists (which decides whether avenger-frontend-developer is provisioned).
  • handover checks codebase/MOC.md staleness.

Keep looking

Skills are one crate of 608,475. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.