Agent Proof Codebase — Testing skill for Claude Code
Skill for Claude Code and opencode: deterministic quality gates (invariant specs, mutation testing, fail-to-pass proof, ratchets, protected referee files) that keep accepted changes high-signal even f.
How to install Agent Proof Codebase
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open mansourvery-hub/agent-proof-codebase and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Agent Proof Codebase does
Skill for Claude Code and opencode: deterministic quality gates (invariant specs, mutation testing, fail-to-pass proof, ratchets, protected referee files) that keep accepted changes high-signal even from weak agents
Alternatives in Testing
- Testing Strategy — This project is small, runs in a terminal, and is mostly deterministic 10.9k ★
- Darwin Skill — 达尔文.skill —— 一个让你的Skill无限进化的系统:评估→改进→测试→保留或回滚 Autoresearch-inspired autonomous skill optimization for Claude C 5.7k ★
- Spec Kit Zh — 💫 帮助你快速上手规格驱动开发的工具包💫 Toolkit to help you get started with Spec-Driven Development Works seamlessly across al 238 ★
README
agent-proof-codebase (skill for Claude Code and opencode)
Generic skill for building, auditing and working inside codebases whose deterministic gates make every accepted change high-signal, even when written by weak or sloppy agents. Entry point: `SKILL.md`.
Install
Pick one location (folder name must stay `agent-proof-codebase`):
- Claude Code, all projects:
~/.claude/skills/agent-proof-codebase/ - Claude Code, one project:
/.claude/skills/agent-proof-codebase/ - opencode, all projects:
~/.config/opencode/skills/agent-proof-codebase/ - opencode, one project:
/.opencode/skills/agent-proof-codebase/(opencode also reads the.claude/skillslocations)
unzip agent-proof-codebase.zip -d ~/.claude/skills/ # or ~/.config/opencode/skills/
If you publish the folder in a git repo you can also install it with the skills CLI you already use.
Use
Say what you want in plain words; the description triggers on requests like:
- "make this repo bulletproof / release-ready", "set up quality gates", "add mutation testing"
- "audit how strong our tests really are", "write invariants for this module"
- "stop agents from gaming the tests" Or invoke explicitly, for example: "Use the agent-proof-codebase skill to audit this repo (mode B)." For an open-ended run: "Use the agent-proof-codebase skill. Orient yourself, write your own plan, then build the gates you judge most valuable. Decide for yourself; ask me only about product or irreversible decisions."
Layout
SKILL.md modes, conduct rules, workflows, quick reference
references/ 13 deep-dive docs, read on demand
scripts/ 9 gate building blocks (Python stdlib / bash), tested
assets/ templates: SPEC, GATES, PR template, CODEOWNERS, CI skeleton, gates.sh, AGENTS snippet...
`scripts/gates.sh` is not shipped: copy `assets/gates.sh.example` into the target repo as `scripts/gates.sh` and adapt the TODO lines to the project's real tool
Related Skills
Cia
Claude Code / Codex skill: audits a codebase as a Viable System (Stafford Beer VSM). Maps modules onto Systems
Scenario Validator Arken
Validate every scenario and edge case in artifacts/qa-validation/qa-release-scenarios.md against the Arken cod
OpenPitStop
A referee your AI coding agent can't cheat. PitStop scans your repo, seals the evidence, catches agents faking
Manual Test
Propose a trivial, reversible edit to visually check the float's minimize/restore around a diff. Cleans itself
Gtm Skills
Reusable AI skills for go-to-market engineering: ICP research, brand voice, account sourcing and scoring, sign
Cc Tools
by Josh Symonds - High-performance Go implementation of Claude Code hooks and utilities. Provides smart lintin
Related Agents
Certora Mutation Testing
Takes a Certora configuration and invariant suite, generates mutation campaigns correctly with certoraMutate a
Transcript Proofer
Proof-read an ASR words.json against the media before captions are written. Re-runs ASR on suspect spans and r
State Syncer
Validate pipeline state integrity (transitions, dedup, required fields per status) and mirror to Google Sheets