- Home
- Skills
- Workflow Automation
- Rule Enforcer: make your CLAUDE.md rules binding
Works with the AI Agents you already use
Rule Enforcer: make your CLAUDE.md rules binding
Turns the rules in your CLAUDE.md or AGENTS.md into Claude Code permission rules that actually block or ask, and tells you honestly which rules can't be enforced.
Free
Rule Enforcer: make your CLAUDE.md rules binding
Example session with this skill installed
make my CLAUDE.md rules binding
- Read your context and instructions
- Compiled the rule enforcer
Permission review: project settings
Enforceable
- Always ask before
npm publish→ askBash(npm publish *)
Partly enforceable
- Never force-push → deny
Bash(git push --force *),Bash(git push -f *)
Gap: a flag after the branch name, or a push from a script, slips past. - Don't read
.env→ denyRead(./.env)
Gap: blocks Claude's file tools; a script that opens the file itself can still read it.
Keep in CLAUDE.md
- "Keep pull requests small" — permissions can't measure PR size.
Merge into .claude/settings.json
{"permissions": {"deny": ["Bash(git push --force *)", "Bash(git push -f *)", "Read(./.env)"], "ask": ["Bash(npm publish *)"]}}
May I merge these entries without removing any existing permissions?
Connects securely to your tools. The creator never sees your data.
About this skill
Claude Code treats CLAUDE.md as advice. Write "never force-push" or "don't read .env" and it usually listens, until it doesn't. Permission rules in settings.json are different: Claude Code itself enforces them, before the command runs.
Rule Enforcer reads your CLAUDE.md or AGENTS.md and sorts every rule:
- Enforceable: turned into
denyoraskpermission entries, such asRead(./.env)orEdit(/migrations/**). - Partly enforceable: you get the entry plus an honest note on how it can be bypassed (a flag after the branch name, a script,
sh -c). - Keep in CLAUDE.md: style and behaviour rules no permission can enforce, with the reason.
It shows the exact JSON first and only merges it into your settings after you say yes, without removing anything already there. If two of your rules conflict, it asks which should stay instead of guessing.
Checked against the current Claude Code permissions docs, including the path gotchas (./ vs / vs // vs ~/) and the fact that .claudeignore does nothing.
Free. For rules learned from your corrections, with approval and Codex support, see Brain; to check task specs before handing them to another agent, see Spec Lint.
Permission rules match the command or path Claude writes, so they are a strong guardrail, not a security boundary: a script or sh -c can get around them (use Claude Code's sandbox for a hard wall). Applies to Claude Code; Codex and other agents don't read Claude Code settings. Style, tone and process rules can't be enforced and stay in CLAUDE.md.
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 5 days ago
- Free with an Agensi account
Requires
Permissions
File Scopes
Never opens env, key or credential files; it protects them by path without reading them.