Install
$ agentstack add skill-stealth-engine-skills-agents-md-setup ✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
Verified badge
Passed review? Show it. Paste this badge into your README, it links to the public security report.
Reliability & compatibility
Declared compatibility
Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.
We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.
How agent discovery & health will work →About
agents-md-setup
Make AGENTS.md the single source of truth for agent/project instructions, and point every tool's file at it via symlink. One file, read by all agents — no duplication, no drift.
Why this layout
AGENTS.mdis the emerging cross-tool standard (Cursor, Codex, and others read
it).
- Claude Code reads
CLAUDE.md. You cannot make the harness readAGENTS.md
instead via a prompt or a skill — but a symlink CLAUDE.md → AGENTS.md means Claude opens CLAUDE.md and gets AGENTS.md's content. That is the mechanism; nothing else is needed. (True as of 2026-07: native AGENTS.md support is still an open request — anthropics/claude-code#34235. Re-verify against the changelog before relying on it; if Claude Code gains native AGENTS.md support the symlink becomes optional.)
- Result: edit
AGENTS.mdonly; every agent stays in sync.
Decision: what's already there?
- Neither file → create
AGENTS.mdfrom the template below. - Only
CLAUDE.md(a real file) → make it the source of truth:
``bash git mv CLAUDE.md AGENTS.md # preserve history (or: mv if not tracked) ``
- Only
AGENTS.md→ keep it as-is (good). - Both exist as real files → do NOT blindly overwrite. Compare them; merge
into AGENTS.md (it wins as the canonical file), confirm with the user if the contents differ, then remove the standalone CLAUDE.md.
CLAUDE.mdalready a symlink →AGENTS.md→ already done; just confirm
the target exists (link isn't dangling), then stop.
AGENTS.mdis a symlink →CLAUDE.md(reversed) → flip it soAGENTS.md
is the real file. Do not git mv the symlink onto its own target (creates a broken/self-referencing link): ``bash rm AGENTS.md # remove the symlink (the real content is in CLAUDE.md) git mv CLAUDE.md AGENTS.md # CLAUDE.md becomes the real AGENTS.md ln -s AGENTS.md CLAUDE.md # recreate CLAUDE.md as the symlink git add AGENTS.md CLAUDE.md ``
Create the symlink
ln -s AGENTS.md CLAUDE.md
git add AGENTS.md CLAUDE.md # git stores it as a real symlink (mode 120000)
Verify it committed as a symlink, not a copy:
git ls-files -s CLAUDE.md # mode should be 120000
Optionally point other tools at the same file:
ln -s AGENTS.md GEMINI.md
Most modern agents already read AGENTS.md (and often CLAUDE.md) natively — e.g. GitHub Copilot reads AGENTS.md directly (coding agent, VS Code, CLI) as of Aug 2025, so no copilot-instructions.md symlink is needed there. Only add a tool-specific symlink for older surfaces that don't yet read AGENTS.md:
# Only for legacy Copilot surfaces that predate native AGENTS.md support:
mkdir -p .github && ln -s ../AGENTS.md .github/copilot-instructions.md
Monorepos (Turborepo / workspaces)
Keep a root AGENTS.md for repo-wide conventions (with the CLAUDE.md symlink as above), then add a scoped AGENTS.md in each workspace that needs its own rules — apps//AGENTS.md, packages//AGENTS.md — each with its own sibling symlink created from inside that directory so the target is relative:
cd apps/web && ln -s AGENTS.md CLAUDE.md # CLAUDE.md -> ./AGENTS.md (same dir)
- Claude Code reads the nearest
CLAUDE.mdfor the files it's working on, so
nested files supplement the root rather than replace it. Put repo-wide things at the root; put only the package-specific delta in each workspace — don't duplicate the root.
- Always make each symlink relative to its own folder (`ln -s AGENTS.md
CLAUDE.md while cd`'d into the workspace), never pointing back at the root — relative same-dir links survive checkout/move.
- Turborepo itself doesn't read these files (they're for agents), so **no
turbo.json change is needed**. If you want them surfaced, you can still note the convention in the root AGENTS.md.
README
If there's no README.md, scaffold a minimal one (project name, one-line description, setup/run commands). Keep human-facing docs in README.md and agent instructions in AGENTS.md — don't merge the two.
Windows / CI caveat (important)
Git symlinks need core.symlinks=true (off by default on Windows Git) and, to create them on checkout, Windows also requires either Developer Mode enabled or an elevated/admin shell. Without those they break on Windows checkouts or some CI runners (the symlink lands as a text file containing AGENTS.md). If the project targets Windows, prefer the fallback: a tiny real CLAUDE.md whose entire content is a pointer —
See [AGENTS.md](./AGENTS.md) for project + agent instructions.
— accepting that it must be kept in sync manually (or via a check). Symlink is better everywhere symlinks are reliable.
Don't undo it later
If you see CLAUDE.md as a symlink to AGENTS.md, that's intentional — edit AGENTS.md, never replace the symlink with a copy.
AGENTS.md template
# AGENTS.md
Instructions for AI agents and humans working in this repository.
`CLAUDE.md` is a symlink to this file; this file is the source of truth.
## Project overview
## Commands
- install: `…`
- dev: `…`
- build: `…`
- test / lint / format: `…`
## Conventions
-
- Commit messages:
-
## Layout
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: stealth-engine
- Source: stealth-engine/skills
- License: MIT
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet, be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.