Agent skill

codemap

Analyze codebase structure, dependencies, changes, cross-agent handoffs, and get code-aware intelligence. Use when user asks about project structure, where code is located, how files connect, what changed, how to resume work, before starting any coding task, when you need risk analysis and skill guidance, or when Codemap should tune project config before analysis.

Stars 507
Forks 45

Install this agent skill to your Project

npx add-skill https://github.com/JordanCoin/codemap/tree/main/plugins/codemap/skills/codemap

SKILL.md

Codemap

Codemap gives you instant architectural context about any codebase. It classifies your intent, detects risk, matches relevant skills, and tracks your working set — all automatically via hooks.

Codemap should also keep its own per-project config healthy. On first use in a repo, or when output is obviously noisy, tune .codemap/config.json before doing deeper analysis so future calls stay code-first instead of asset-first.

Commands

bash
codemap .                       # Project structure and top files
codemap --deps                  # Dependency flow (imports/functions/hubs)
codemap --diff                  # Changes vs main branch
codemap --diff --ref <branch>   # Changes vs specific branch
codemap --importers <file>      # Who imports this file? Is it a hub?
codemap handoff .               # Build + save handoff artifact
codemap handoff --latest .      # Read latest saved handoff
codemap handoff --json .        # Machine-readable handoff payload
codemap skill list              # Show available skills with descriptions
codemap skill show <name>       # Get full skill instructions
codemap skill init              # Create custom skill template
codemap config show             # Show current project config
codemap context                 # Universal JSON context envelope
codemap context --for "prompt"  # With pre-classified intent + matched skills
codemap context --compact       # Minimal for token-constrained agents
codemap serve --port 9471       # HTTP API for non-MCP integrations

First-Use Setup

Before deeper Codemap analysis in a repo:

  1. Check .codemap/config.json.
  2. If it is missing, clearly boilerplate, or obviously too noisy for the stack, run codemap skill show config-setup and follow it.
  3. After writing or improving config, rerun codemap . and codemap --deps.

Treat config as repo memory. Once tuned, future Codemap calls should benefit automatically.

Signals that config needs setup or tuning:

  • .codemap/config.json is missing
  • config only contains generic auto-detected only values with no real project shaping
  • large non-code directories dominate the tree output
  • stack-specific noise is overwhelming source structure (.xcassets, screenshots, PDFs, training-data, fixtures, generated files, models, vendor directories)
  • the repo stack is obvious, but the config does not reflect it

When to Use

ALWAYS run codemap . when:

  • Starting any new task or feature
  • User asks "where is X?" or "what files handle Y?"
  • User asks about project structure or organization
  • You need to understand the codebase before making changes

ALWAYS run codemap --deps when:

  • User asks "how does X work?" or "what uses Y?"
  • Refactoring or moving code
  • Need to trace imports or dependencies
  • Finding hub files (most-imported)

ALWAYS run codemap --diff when:

  • User asks "what changed?" or "what did I modify?"
  • Reviewing changes before commit
  • Summarizing work done on a branch

ALWAYS run codemap --importers <file> when:

  • About to edit a file — check if it's a hub
  • Need to know the blast radius of a change
  • Deciding whether to refactor or leave alone

Run codemap skill show <name> when:

  • The prompt-submit hook shows matched skills in <!-- codemap:skills [...] -->
  • You need guidance for a specific task (hub editing, refactoring, testing)
  • Risk level is medium or high

Run codemap skill show config-setup when:

  • The repo has no .codemap/config.json
  • The config looks like a bare bootstrap and not a real project policy
  • Codemap output is cluttered by large non-code directories
  • You want Codemap to make better future decisions for this specific repo

Run codemap context when:

  • Piping codemap intelligence to another tool
  • Need a structured JSON summary of the project state
  • Building automation that consumes code-aware context

Run codemap handoff when:

  • Switching between agents (Claude, Codex, Cursor)
  • Resuming work after a break
  • User asks "what should the next agent know?"

Hook Output

The prompt-submit hook fires on every message and provides:

<!-- codemap:intent {"category":"refactor","risk":"high",...} -->
<!-- codemap:skills [{"name":"hub-safety","score":5},...] -->
Skills matched: hub-safety, refactor — run `codemap skill show <name>` for guidance
  • Intent categories: refactor, bugfix, feature, explore, test, docs
  • Risk levels: low (no hubs), medium (1 hub), high (2+ hubs or 8+ importers)
  • Skills are pull-based: only names are shown, run codemap skill show for full body

Builtin Skills

Skill When to Pull
config-setup Missing, boilerplate, or noisy .codemap/config.json
hub-safety Editing files imported by 3+ others
refactor Restructuring, renaming, moving code
test-first Writing tests, TDD workflows
explore Understanding how code works
handoff Switching between AI agents

Output Interpretation

Tree View (codemap .)

  • Stars (⭐) indicate top 5 largest source files
  • Directories flattened when containing single subdirectory

Dependency Flow (codemap --deps)

  • External dependencies grouped by language
  • Internal import chains showing how files connect
  • HUBS section shows most-imported files (3+ importers)

Diff Mode (codemap --diff)

  • (new) = untracked, = modified, (+N -M) = lines changed
  • Warning icons show hub files (high impact)

Importers (codemap --importers <file>)

  • Shows all files that import this file
  • Flags hub status (3+ importers = high impact)

Context Envelope (codemap context)

  • JSON with project metadata, intent, working set, matched skills, handoff ref
  • --compact strips skills and limits working set for token savings

MCP Tools

If codemap MCP server is configured, these tools are available:

Tool Use For
get_structure Project tree
get_dependencies Dependency flow + hubs
get_diff Changed files with impact
find_file Search by filename
get_importers Who imports a file
get_hubs List all hub files
get_file_context Full context for one file
get_handoff Build/read handoff artifact
get_working_set Files edited this session
list_skills Available skills (metadata)
get_skill Full skill instructions
get_activity Recent coding activity
start_watch / stop_watch Control daemon
status Verify MCP connection
list_projects Discover projects

HTTP API

When codemap serve is running:

Endpoint Returns
GET /api/context?intent=... Context envelope
GET /api/skills All skills metadata
GET /api/skills/<name> Full skill body
GET /api/working-set Current working set
GET /api/health Server health

Expand your agent's capabilities with these related and highly-rated skills.

JordanCoin/codemap

config-setup

Set up or tune .codemap/config.json so Codemap focuses on code-relevant parts of the repo. Use when config is missing, boilerplate, noisy, or mismatched to the stack.

507 45
Explore
JordanCoin/codemap

codemap

Analyze codebase structure, dependencies, changes, cross-agent handoffs, and get code-aware intelligence. Use when user asks about project structure, where code is located, how files connect, what changed, how to resume work, before starting any coding task, when you need risk analysis and skill guidance, or when Codemap should tune project config before analysis.

507 45
Explore
JordanCoin/codemap

config-setup

Set up or tune .codemap/config.json so Codemap focuses on code-relevant parts of the repo. Use when config is missing, boilerplate, noisy, or mismatched to the stack.

507 45
Explore
davila7/claude-code-templates

verl-rl-training

Provides guidance for training LLMs with reinforcement learning using verl (Volcano Engine RL). Use when implementing RLHF, GRPO, PPO, or other RL algorithms for LLM post-training at scale with flexible infrastructure backends.

23,776 2,298
Explore
davila7/claude-code-templates

openrlhf-training

High-performance RLHF framework with Ray+vLLM acceleration. Use for PPO, GRPO, RLOO, DPO training of large models (7B-70B+). Built on Ray, vLLM, ZeRO-3. 2× faster than DeepSpeedChat with distributed architecture and GPU resource sharing.

23,776 2,298
Explore
davila7/claude-code-templates

gguf-quantization

GGUF format and llama.cpp quantization for efficient CPU/GPU inference. Use when deploying models on consumer hardware, Apple Silicon, or when needing flexible quantization from 2-8 bit without GPU requirements.

23,776 2,298
Explore

Didn't find tool you were looking for?

Be as detailed as possible for better results