Works with the AI Agents you already use
reforge
Audit and clean up messy, vibe-coded repos: dead code, duplicates, god files and circular deps, fixed one safe commit at a time. Requires the free reforge-mcp server.
Free
reforge
Example session with this skill installed
Scan the current repository, give me a health score, and find the three highest-impact dead code symbols to remove.
- Read your context and instructions
- Compiled the reforge
I have completed the initial scan and risk assessment of the repository. The scanner successfully parsed 84% of the source files, with 12 files skipped due to unsupported extensions.
Repository Audit Results
| Metric | Value | Notes |
|---|---|---|
| Initial Health Score | 64/100 | Coverage: 0.84 |
| Architecture Pattern | Modular Monolith | Entry point: src/index.ts |
| Parse Failures | 2 | src/legacy/parser.py (SyntaxError) |
| Total Issues Found | 22 | Weighted by score_risk |
Top Priority Fixes
src/utils/old_helpers.ts:formatLegacyDate— Dead Code (Risk: 0.1)src/auth/session.ts:validateTokenV1— Duplicated Logic (Risk: 0.2)src/core/engine.ts:InternalBuffer— Unused Export (Risk: 0.2)
I have saved these to pending_work in the repository memory.
Next steps
- Say "fix the first item" and I'll apply the removal of formatLegacyDate as a single commit.
- Review the
architecture_hypothesisstored in memory for module boundaries. - Address the two
parse_failuresin the legacy directory to improve audit coverage.
Connects securely to your tools. The creator never sees your data.
About this skill
⚠️ Requires a local coding agent + reforge-mcp. Works in Claude Code, Cursor, Codex CLI and other agents that run local MCP servers. It does not work in the Claude web app or ChatGPT, because reforge needs direct access to your repository.
Setup (2 minutes)
- Install the free, open-source MCP server:
pip install reforge-mcp - Register it with Claude Code:
claude mcp add reforge -- reforge-mcp - Add this skill, then ask your agent to scan your repo.
Source: github.com/Codewithaiyan/reforge-mcp
What it does
This skill teaches your agent to clean up a repository using real tools instead of guessing. It parses your code, scores its health, ranks problems by risk, and fixes them one at a time with a git commit per fix, so every change is easy to review or revert.
- Health score: a 0–100 baseline, tracked over time
- Architecture inference: entry points, modules and project pattern
- Finds: dead code, duplicated logic, god files, circular dependencies
- Risk ranking: low-risk, high-payoff fixes first
- Safe fixes: each fix is applied as a diff, can run your tests, and rolls back automatically if they fail
- Changelog: a
REFORGE_CHANGES.mdsummarizing every change and the score delta
How it works
- Scan & score:
scan_repoandget_health_score - Plan: rank findings and save the plan with
write_memory, so it survives restarts - Fix: read only the needed lines with
get_chunk, apply one fix withwrite_fix - Commit & report: one commit per fix with
git_commit, thengenerate_changelog
The agent stops and asks you at confirmation checkpoints and respects a per-session commit budget.
Why not just prompt it?
A plain prompt reads a few files and guesses. reforge parses the whole repo first, so fixes are based on what is actually used, not on what looks unused.
Only analyzes git-tracked files. Supports Python, JavaScript, TypeScript and Go. Dead-code detection is name-based, so it can flag framework callbacks (routes, hooks, decorators) as unused; review these before accepting removal..
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 4 days ago
- Free with an Agensi account
Requires
Permissions
Reads and edits files only inside the repository you point it at, and creates git commits there. The reforge-mcp server makes no network calls.