repo-intake

Repo Intake

When to Use

• First time onboarding a project
• Rebuilding cache after major project structural changes
• Cache expired, needs updating

When NOT to Use

• Already familiar with project structure (read cache directly)
• Only need to find specific files (use Glob/Grep)
• Day-to-day development (cache already exists)

Workflow

Docs -> Entrypoints -> Tests Map -> Next Steps

Usage

bash scripts/run-skill.sh repo-intake intake_cached.js --mode auto --top 10

Manifest Map (declared dependency map)

Workspace topology anddeclareddependency graph for monorepo / multi-package repos. Edge semantics is declares_dependency—what is written in the manifest does not prove import, call, or runtime impact. Always fail-closed: if the entire file cannot be read / UTF-8 decoded, or the JSON document cannot be parsed, record coverage skipped; unsupported constructs that have entered the Go/TOML recognizer are recorded as partial. Never generate speculative edges.

bash scripts/run-skill.sh repo-intake manifest_map.js                      # overview (md)
bash scripts/run-skill.sh repo-intake manifest_map.js --format json        # 完整 envelope
bash scripts/run-skill.sh repo-intake manifest_map.js --reverse <selector> # 誰宣告依賴它
bash scripts/run-skill.sh repo-intake manifest_map.js --cycles             # 宣告環偵測

Flag	Description
--format md|json	Default md; json is the complete artifact, not subject to --top truncation
--top N	md list truncation (default 12)
--reverse <sel>	selector: node ID (ws:node:packages/a) or name (node:lodash / lodash); multiple matches exit 2 and list candidates
--cycles	and --reverse are mutually exclusive
--include-candidates	Include candidate workspaces not confirmed by the controller into the architecture node set (excluded by default)

Ecosystems: node / php / go / rust / python (for recognized manifests see scripts/config/repo-intake.json the manifest_map section). Invalid argument exit 2. Data model, parsing matrix, and frozen syntax contract:references/manifest-map.md; full specification: docs/features/repo-intake-manifest-map/2-tech-spec.md。

Cache Location

Cache stored at: ~/.claude/cache/repo-intake/<repoKey>/

File	Description
latest.md	Latest scan results
latest.json	Latest scan results (JSON)
LATEST.json	Cache metadata

Output

## Overview

<summary>

## Entrypoints

- {CONFIG_FILE}
- {BOOTSTRAP_FILE}

## Test Map

| Type        | Pattern           |
| ----------- | ----------------- |
| Unit        | test/unit/        |
| Integration | test/integration/ |
| E2E         | test/e2e/         |

## Next Steps

- <questions>

Verification

• Output includes Overview, Entrypoints, Test Map, Next Steps
• Entrypoints correctly identify {CONFIG_FILE}, {BOOTSTRAP_FILE}
• Test Map covers Unit/Integration/E2E layers

References

• references/manifest-map.md — Manifest map data model, parsing matrix, frozen contract summary
• references/archived/MIDWAY_HEURISTICS.md — Legacy MidwayJS heuristics (archived, for reference only)

Scripts

Script	Purpose
scripts/intake_cached.js	Main intake with caching
scripts/scan_repo.js	Full repo scanner (framework-agnostic)
scripts/scan_delta.js	Delta scan for changed files
scripts/manifest_map.js	Workspace topology + declared dependency graph (overview / reverse / cycles)

Examples

Input: /repo-intake
Action: Execute intake script -> Output project map

Input: /repo-intake save
Action: Execute intake script -> Output and write to docs/ai/intake/