Rulewitness — Development skill for Claude Code
Semantic permission diffs and policy tests for Claude Code.
How to install Rulewitness
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open vlovers/rulewitness and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Rulewitness does
Semantic permission diffs and policy tests for Claude Code. Explain permission changes with concrete witnesses.
Alternatives in Development
- 11 Permission Explainer — Permission Explainer System Prompt 2.3k ★
- Phi — a coding Agent from pi 175 ★
- Generate Permissions — Generate VAMS Permission Template 137 ★
README
RuleWitness
**Semantic Permission Diff for AI Coding Agents**
[](https://github.com/vlovers/rulewitness/actions/workflows/ci.yml?query=branch%3Amain) [](LICENSE) [](https://www.npmjs.com/package/rulewitness)
**Did a pull request change what Claude Code can do?** RuleWitness compares supported permission rules between Git revisions, shows a concrete command when a modeled decision changes, and explains which rules caused it. It also detects modeled no-ops and supports policy assertions — offline, without executing the commands.

RuleWitness is an experimental, offline TypeScript CLI and library for reviewing **Claude Code configuration-rule decisions** between Git commits. It produces concrete, replay-checked witnesses, explains the determining rules, filters modeled no-ops, and runs reusable policy assertions.
**This is not a runtime authorization engine.** `ALLOW` means the supported configured rules allow that input in this model. It does not guarantee Claude Code will execute it. Every report retains `complete: false`. Unsupported settings and uncertain changes remain visible. Only Claude Code's documented subset below is supported; other agents are not advertised.
This project evolved from AgentPermissionDiff in v0.2. Version v0.3 adds a reviewer-focused Markdown presentation, without expanding the permission model. The repository is [vlovers/rulewitness](https://github.com/vlovers/rulewitness), and `rulewitness@0.3.0` is [published on npm](https://www.npmjs.com/package/rulewitness/v/0.3.0). See [migration notes](docs/migration-v0.2.md), the [v0.2.0 release](https://github.com/vlovers/rulewitness/releases/tag/v0.2.0), and the [historical release preparation notes](doc
Related Skills
Explain Code Changes
Review all local code changes with detailed diffs and explanations
A2acode
Serve any ACP coding agent — Claude Code, Gemini CLI, Codex — over the A2A protocol, with tool progress, file
Explain Error
You are executing the /explain-error command. Your job is to diagnose a Coq type error: parse it, gather conte
Explain Diff HTML
Turn pull requests and code diffs into interactive HTML walkthroughs that explain the logical flow of a change
Suggest Changes
Analyze the agent-monitor debug logs and codebase to discover concrete improvements, fixes, and optimizations.
Lightclaw
Control local Codex and Claude coding agents from your phone via Telegram — review plans, approve bounded work
Related Agents
Legal Compliance Reviewer
Use before publishing a web or mobile app, or after changes to what personal data it collects, logs, shares or
API Sweeper
Explain canonical Sentinel 2.0 API and RBAC findings without making requests or policy decisions
Permission Analyst
Recommends allow/deny permission rules from denial telemetry and policy via PromptWise. Use to reduce permissi