Works with the AI Agents you already use
stackfit: only the agent skills each repo needs
Sets up stackfit so each repo gets only the skills, rules, and commands that fit its stack, for Claude Code, Cursor, and Codex, with an honest before/after context report.
Free
stackfit: only the agent skills each repo needs
Example session with this skill installed
Set stackfit up for my repos in ~/code and tell me honestly how much context I save compared with today.
- Read your context and instructions
- Compiled the stackfit
stackfit is set up. Your catalog at ~/.stackfit has only your own 3 skills: core (every repo): our-api-conventions, grilling; react (repos using react): vercel-react-best-practices. Both repos are synced and up to date; your old global copies are in ~/.claude/skills-backup/.
Honest savings: today all 3 skills load in every Claude Code session (~153 tokens). web-app still loads ~153 (it needs all three, so no saving). worker drops to ~64 (about 58% less). In absolute terms the saving is small because you only had 3 skills; the main gain is one catalog and React guidance staying out of the Go repo. Note: Cursor and Codex now see these skills in both repos, which they didn't before.
Connects securely to your tools. The creator never sees your data.
About this skill
Every installed skill puts its name and description into every agent session. Install 30 skills globally and a Go service carries your React animation guides on every turn.
This skill teaches your agent to run stackfit, a free, open-source CLI (MIT, zero dependencies): one catalog of skills, split into packs that are picked per repo from its manifests (package.json, requirements.txt, pyproject.toml, go.mod, Cargo.toml, composer.json), and rendered natively for Claude Code, Cursor, and Codex across all your projects.
What your agent does with it:
- Measures what loads today (global skill folders per tool) before changing anything.
- Lets you choose: organise only your own skills, or start from a curated starter (Vercel, shadcn, Anthropic, Stripe, Emil Kowalski, Matt Pocock skills).
- Previews every change, never overwrites your own files, and moves your existing skills into the right place.
- Reports honestly per repo: skill listings + always-on rules per session, against your global baseline. If a repo would load more than today, it says so.
- Handles per-repo choices, several machines, upstream refreshes with your edits preserved, team catalogs, and CI checks.
Every command, recipe, and error message in the skill was checked against the real CLI and blind-tested by fresh agents. Source and demo: https://github.com/amin0075/stackfit
Requires the free stackfit CLI (npx stackfit, Node 22+). Token figures are estimates (about 4 characters per token) of always-loaded skill listings and rules, not total usage. Codex does not read rule files. Skills in an always pack reach every synced repo, not sessions outside your project folders. Vendoring upstream skills needs git access to GitHub.
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 forever
No account required to browse
Trust & safety
Security scan
Security scan passed 21 days ago
- Free with an Agensi account
Requires
Permissions
Allowed Hosts
File Scopes
Never commits, pushes or uses --force without your explicit OK. Personal values stay in a gitignored .env.