By annyeong844
Audits TypeScript and JavaScript repos for structural debt—dead exports, cycles, oversized modules, duplicate helpers, naming drift—and generates evidence-backed refactor plans. Guards edits with pre-write reuse checks and post-write delta validations to prevent unintended changes.
Default Lumin Repo Lens entrypoint. With no arguments, run a baseline-aware current-workspace structure review; do not ask the user to choose a mode.
Run the post-write delta check after edits. Usage: /lumin-repo-lens:post-write --root <repo> --pre-write-advisory <pre-write-advisory.<invocationId>.json> [--output <dir>]
Run the pre-write reuse gate before editing code. Give a natural-language change request, or pass --intent <file|-> for machine use.
Turn repo evidence into a kind, incremental refactoring plan. Usage: /lumin-repo-lens:refactor-plan [--root <repo>] [--output <dir>] [--profile quick|full]
First-touch guide for Lumin Repo Lens. Use when the user is new, unsure what to run, or wants a gentle explanation before reviewing repo structure.
Maintainer-only canon surface for Lumin Repo Lens/lumin-repo-lens: draft or check promoted repository facts, canonical drift, canon-draft/check-canon results, parser contracts, and maintainer validation evidence.
Use before/after TS/JS code changes: add, edit, move, rename, refactor, make a helper/type/file/function, or ask if something already exists. Infer intent from plain language, run pre-write reuse screening, then post-write delta checks.
Audit TypeScript/JavaScript repos for structural debt with machine evidence: dead exports, cycles, oversized modules, duplicate helpers/types, barrels, naming drift, refactor plans, 'what should I clean up next?', and 'does X exist anywhere?'
Matches all tools
Hooks run on every tool call, not just specific ones
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
🇰🇷 한국어로 읽으시려면 → README.ko.md · 🇬🇧 English continues below.
The kind little buddy that says "this already exists" before you write it again. Your repo's companion for vibe-coding sessions.
You ask AI:
"Make me a card-news service"
AI happily spins up a new file → lib/cardNewsService.js ✨
But your repo already had three similar files in lib/cardNews/. 😱
With lumin-repo-lens riding along, Claude can say first:
🛑 "Wait — there are already 3 similar files under
lib/cardNews*. Want to look before making a new one?"
Real evidence from your actual repo, not guesses.
1. Add the marketplace and install the plugin in Claude Code
/plugin marketplace add annyeong844/lumin-repo-lens
/plugin install lumin-repo-lens@annyeong844-marketplace
/reload-plugins
2. Run the full first checkup
/lumin-repo-lens:full
→ This turns on the parts that make Lumin more than a dead-export sorter: shape index, function-clone cues, call graph, barrel discipline, topology, public-surface policies, and the grounded review profile.
3. Use the write gate when you are about to change code
/lumin-repo-lens:pre-write
# code the change
/lumin-repo-lens:post-write
Claude Code infers the compact intent internally. You do not need to write the intent JSON by hand.
💡 First run installs parser dependencies once (~30 seconds). After that, fast. For tiny follow-up checks after a fresh baseline,
/lumin-repo-lensuses the quick path.
For very large repos, do not auto-trigger full profile on every edit. Run
:full once per branch, first checkup, or major refactor review, then use
pre-write/post-write and quick follow-ups during the agent loop.
| Command | When to use |
|---|---|
/lumin-repo-lens:full | Full evidence profile — first checkup, post-refactor review, shape/function-clone/call/topology evidence |
/lumin-repo-lens:pre-write | Check before coding. Ask naturally; the assistant infers the compact intent internally. |
/lumin-repo-lens:post-write | Verify after coding — "did my change ripple anywhere else?" |
/lumin-repo-lens | Quick baseline-aware repo lens pass for small follow-up checks over fresh artifacts |
/lumin-repo-lens:refactor-plan | Turn evidence into a cautious cleanup plan |
/lumin-repo-lens:welcome | Get a gentle first-use menu |
Maintainers can also use /lumin-repo-lens:canon-draft and
/lumin-repo-lens:check-canon for canon lifecycle work.
First pass, stale or missing artifacts, explicit review, due diligence,
large refactor planning, and post-refactor review should run --profile full.
Small follow-up checks over a fresh baseline can use the quick path.
In the generated skill package, use the public wrapper:
git clone https://github.com/annyeong844/lumin-repo-lens.git
cd lumin-repo-lens
node skills/lumin-repo-lens/scripts/audit-repo.mjs --root <repo>
LUMIN_REPO_LENS_NO_AUTO_INSTALL=1 node skills/lumin-repo-lens/scripts/audit-repo.mjs --root <repo>
With this set, the tool prints the exact install command instead of running it for you.
The automatic setup command is npm ci --omit=dev --ignore-scripts --no-audit --fund=false.
Stable validation modes are audit, pre-write, post-write, canon-draft,
and check-canon.
Codex users can use the $lumin-repo-lens-codex wrapper, which points at the shared engine.
git clone https://github.com/annyeong844/lumin-repo-lens.git ~/.codex/lumin-repo-lens
macOS / Linux
mkdir -p ~/.codex/skills
ln -sfn ~/.codex/lumin-repo-lens/skills/lumin-repo-lens-codex ~/.codex/skills/lumin-repo-lens-codex
ln -sfn ~/.codex/lumin-repo-lens/skills/lumin-repo-lens ~/.codex/skills/lumin-repo-lens
ln -sfn ~/.codex/lumin-repo-lens/skills/lumin-repo-lens-write-gate ~/.codex/skills/lumin-repo-lens-write-gate
ln -sfn ~/.codex/lumin-repo-lens/skills/lumin-repo-lens-canon ~/.codex/skills/lumin-repo-lens-canon
Windows PowerShell
npx claudepluginhub annyeong844/lumin-repo-lens --plugin lumin-repo-lensLumin Repo Lens Lab: isolated performance and Rust sidecar experiments.
Personal Claude Code + Codex dev stack: security hooks, AI-first code conventions, /security-review, /repo-map, /stack-check, portable statusline. Designed to complement other skills-based plugins, not replace them.
Codebase exploration, refactoring, and quality analysis
Local codebase intelligence + change-safety gates for coding agents. Pre-indexes your repo's symbols, call graph, deps, and git history into SQLite (28 languages, 100% local, zero API keys), then exposes a lean 16-tool MCP core preset: graph-precise search, callers/impact blast radius, coupling, dead code, taint reachability, and pre-merge verify/critique gates that catch regressions, broken references, AI-slop duplication, and convention drift before they ship.
AI-powered codebase understanding assistant. Learn design patterns, analyze impact, trace code flows, and understand any codebase through information theory principles. Includes 6 Agent Skills for automatic analysis triggering.
Dead code analysis, duplication detection, complexity hotspots, and auto-fix for JavaScript/TypeScript using fallow
Live codebase visualization and structural quality gate — 14 health dimensions graded A-F, dependency analysis, and architecture governance via MCP