- Home
- Skills
- Agents & Orchestration
- holdfast
holdfast
Holdfast helps Claude Code enforce persistent, plain-English rules through tool-call hooks, preventing AI agents from violating user-defined constraints during coding tasks.
Free
Works with the AI Agents you already use
holdfast
Example session with this skill installed
Install Holdfast in this repo and make sure I don't delete any files in the 'docs' folder without asking me first.
- Read your context and instructions
- Compiled the holdfast
I have installed Holdfast and registered your protection rule. Claude Code will now prompt for your manual approval before any tool attempts to modify or delete files within the docs/ directory.
Installation and Rule Status
| Component | Status | Details |
|---|---|---|
| Python Environment | Installed | ~/.holdfast/venv |
| Hook Status | Active | Registered in .claude/settings.local.json |
| Active Rules | 1 Found | "Don't delete any files in the 'docs' folder without asking me first." |
| Last Decision | N/A | No actions attempted since install |
Holdfast is now monitoring all tool calls. If I attempt to run a command like rm docs/readme.md or a Python script that deletes files in that path, Holdfast will intercept the call and wait for your explicit "yes" in the chat.
Next steps
- Run
git statusto trigger a harmless tool call and confirm the hook is firing. - Check
~/.holdfast/venv/bin/holdfast statusto see the currently active session rules. - Try a dry run with
~/.holdfast/venv/bin/holdfast explain --rule "Don't delete docs" -- "rm docs/test.txt"to see the guard in action.
Connects securely to your tools. The creator never sees your data.
About this skill
Holdfast helps developers keep AI coding agents under control.
- Enforces plain-English rules through Claude Code hooks.
- Blocks actions that violate user-defined constraints.
- Keeps rules active across long sessions.
- Helps prevent unwanted code changes.
Requires Holdfast CLI and Claude Code.
Requires macOS/Linux and Python 3.10+. Not a security sandbox. 63-92ms overhead per tool call. Effectiveness depends on clear phrasing of rules.
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 6 days ago
- Free with an Agensi account
Requires