- Home
- Skills
- Prompt & Skill Engineering
- Project-Specific AGENTS.md Generator
Project-Specific AGENTS.md Generator
Preserve existing policy, label proposed defaults and distinguish configured commands from checks actually run.
$7
Works with the AI Agents you already use
Project-Specific AGENTS.md Generator
Example session with this skill installed
Use agents-md-generator to draft AGENTS.md for the real example-app project inside public harness snapshot 96f9fcee9453c6bec98026f40f76ed426bad3e6e. Save the proposed file here, without changing the source repo. Preserve inherited public-repository safety boundaries. Do not introduce a second agent, branching or commit procedure as if required by the example app. No dependencies or code execution.
Input: ../_sources/harness/example-app, inherited ../_sources/harness/AGENTS.md and CLAUDE.md. Codex followed the skill in this session, read manifest and source files, and verified the document statically. Source repo remains unchanged. Generated document is a proposal; existing inherited instructions would still govern any later work inside the real repo.
- Read your context and instructions
- Compiled the project-specific agents md
These are agent-authored outputs from a real review of the labeled inputs, not a bundled script's automatic result, independent validation or a promised future outcome.
AGENTS.md — example-app
Proposed project-specific companion to the repository-root instructions. Those inherited instructions still apply; this file does not override their orchestration or publication rules.
Project scope
This is a TypeScript/Express task API demo. It uses an in-memory task store. Keep changes within example-app unless the task explicitly involves another part of the repository. The dashboard is a separate project; do not apply its build commands here.
Files to inspect
- src/app.ts: application factory and route/error middleware.
- src/server.ts: server entrypoint.
- src/routes/tasks.ts: task endpoints.
- src/store.ts: in-memory Map and task operations.
- src/validation.ts: Zod input schemas and task types.
- tests/: Vitest/Supertest tests.
Configured commands
Run from example-app, after dependencies are available through an authorized setup:
- npm test — configured as vitest run.
- npm run typecheck — configured as tsc --noEmit.
- npm run dev — configured as tsx src/server.ts.
- npm start — configured as tsx src/server.ts.
No build script is defined in this package. Commands above were found in package.json; they were not executed during generation. Do not claim the project passes checks based on this document.
Work and verification
Follow the inherited root workflow for real changes. Inspect the endpoint and validation/store behavior before editing. For API behavior changes, use the applicable tests and typecheck; report actual output and explain failures. Keep test stores isolated. Avoid installing dependencies or starting a server unless needed and authorized for the task.
Boundaries and conventions
Keep public content in English. Do not include private data, credentials, local machine paths, session links or personal images. Do not read outside the project scope. Show changes before committing and never push without the user's explicit authorization, as required by root instructions. Do not modify dashboard/.data or use real user records in tests.
Keep ESM-style imports consistent with existing source and respect strict schema validation. Task titles and descriptions are free text; keep fixtures synthetic. Do not silently replace the in-memory store with persistent storage or expand demo scope.
Completion report
Explain the resulting behavior, changed files, checks actually run, failures/limitations and remaining work. Do not assert a test count from documentation as a fresh result.
Generation evidence and limits
Run on real public harness example-app, snapshot 96f9fcee9453c6bec98026f40f76ed426bad3e6e, 2026-11-14. Full output: AGENTS.generated.md.
| Output fact/rule | Evidence |
|---|---|
| Express factory and middleware | example-app/src/app.ts:1-23 |
| In-memory Map | example-app/src/store.ts:6-7 |
| Task schemas and types | example-app/src/validation.ts:1-36 |
| Four configured commands and absence of build | example-app/package.json:6-11 |
| Source/test paths | actual file inventory saved in static-check.json |
| English, public-data rules, no external reads, no push, diff before commit | root AGENTS.md:12-20; CLAUDE.md:3-10 |
| Preserve dashboard data and isolated tests | root AGENTS.md:20 and technical-pitfall guidance |
| No new persistence and no invented test results | proposed scope-preserving guidance derived from the in-memory demo; not a claim of an existing project rule |
No package scripts ran; dependency presence and runtime behavior were not verified. The proposal supplements rather than replaces inherited instructions. Tool names npm/tsx/Vitest/Zod derive from the real manifest and source. No fictional build command, test count or release status added.
Connects securely to your tools. The creator never sees your data.
About this skill
Read existing/inherited instructions, manifests and representative source, then draft minimal project-specific agent guidance. Preserve existing policy, label proposed defaults and distinguish configured commands from checks actually run. Includes SKILL.md only. When both instruction files are requested, the workflow can keep one canonical source and a short companion. It does not run builds, set global configuration, commit or push as part of generation.
Does not execute scripts to verify them. Static analysis only. Limited to what is visible in the provided repository files. Requires user confirmation for ambiguous policy conflicts.
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
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 2 days ago