Claude Code Guardrails: Safe, Honest Agent Kit
Four Claude Code hooks and tools that stop the most expensive things an AI coding agent does when nobody is watching.
- Proof-of-Done (Claude Code Hooks): the agent cannot end its turn saying "done" until the tests really ran green after its last edit, and deleted, skipped or weakened tests are blocked.
- Safety Guard: blocks rm -rf, git reset --hard, force pushes and DROP TABLE only when they would really lose work (it checks git status first). Bash, PowerShell and cmd.
- Fake Package Blocker: every pip / npm / cargo install is checked for typosquats and hallucinated package names before it runs.
- Memory Saver: your rules, goal, edited files and last test result survive /compact.
Install once, and long autonomous sessions stop going wrong silently. Local only: readable Python scripts, no network by default, no telemetry, each with a visible off switch.
Works with any agent that signs in to Agensi, such as Claude, ChatGPT, Claude Code, Cursor and Codex CLI.
One payment, lifetime access. 4 skills unlock instantly in your library.
What's included
4 skillsYour agent can't end its turn until the tests ran green after its last edit. Also catches deleted, skipped and weakened tests.
Turn /compact from a reset into a checkpoint: your rules, goal, edited files and last test result are saved before compaction and put back after it.
Turn rm -rf, git reset --hard and force pushes into a checked step: blocks destructive shell commands only when they would really lose work.
Turn every pip/npm/cargo install your agent runs into a checked one: blocks typosquats and hallucinated package names before they reach your machine.
See it in action
Claude Code Hooks: Block "Done" Until Tests Really Pass
Claude Code Hooks: Block "Done" Until Tests Really Pass
Example session with this skill installed
Change greet() to return "Hello, <name>!". Just make the edit and tell me it's done; you don't need to run anything.
- Read your context and instructions
- Compiled the claude code hooks
Edit made, but Proof-of-Done blocked the end of the turn:
you changed app/greet.py and no test command has run since your last change. Run
python -m pytest(without piping it into tail/grep), read the result, and only then report.
$ python -m pytest
F
FAILED tests/test_greet.py::test_greet - AssertionError: assert 'Hello, Ada!' == 'Hi Ada'
1 failed
The existing test still asserts the old greeting. I updated it to the behaviour you asked for and re-ran:
$ python -m pytest
.
1 passed in 1.72s
Done. greet() returns "Hello, <name>!".
Evidence: python -m pytest → exit 0, "1 passed", after the last edit.
(Real Claude Code session with Proof-of-Done installed, trimmed.)
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 these skills and uses them on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every skill in this bundle goes through our trust checks.
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.