memgrafter/repo-map

Use this skill when you need quick repo-level context and want a reliable “where to look next” map.

Why use it

• Gives you a high-signal map of files and symbols for planning.
• Helps you choose targets before spending tokens on full file reads.
• Deterministic output makes repeated runs stable and comparable.
• No external aider install required in your environment.

When to use

• Starting work in an unfamiliar repository.
• Scoping a feature, bug, or refactor across many files.
• Finding likely owners of concepts (routes, auth, handlers, models).
• Creating a first-pass plan before detailed inspection.

When NOT to use

• You already know the exact file(s) to edit.
• You need semantic correctness from full code understanding.
• You only need a tiny local change in a well-known area.

Usage

$HOME/.agents/skills/repo-map/run.sh "optional hint" -d <directory> [--map-tokens N] [--json]

Examples

./run.sh "auth" -d .
./run.sh "routing and handlers" -d src --map-tokens 3000
./run.sh "" -d /path/to/repo --json

Output

Returns:

• repo-map: ranked tree-like map text
• file_count: number of scanned source files
• map_tokens: token budget used for map generation
• refresh: refresh policy applied
• error: non-empty if map generation failed

How it works

This skill calls a vendored Aider repomap implementation to build a compact structural map from repository files and optional hint text.

Trade-off: you get fast, broad orientation (benefit) but not full semantic understanding of every code path (cost).