Proof Auditor — Development agent for Claude Code
Independent rubric-verdict producer for reasoning deliverables.
How to install Proof Auditor
Installs to ~/.claude/agents/mememade-github-claude-meta-autoagent-proof-auditor.md
mkdir -p ~/.claude/agents && curl -fsSL https://raw.githubusercontent.com/mememade-github/claude-meta-autoagent/HEAD/.claude/agents/proof-auditor.md -o ~/.claude/agents/mememade-github-claude-meta-autoagent-proof-auditor.md Restart Claude Code, or start a new session, for it to be picked up.
What Proof Auditor does
name: proof-auditor description: Independent rubric-verdict producer for reasoning deliverables. Runs alongside the incumbent judge; produces an independent score and an explicit agreement/disagreement assessment against a prior JUDGMENT. tools: ["Read", "Bash", "Grep", "Glob"] model: opus maxTurns: 30 color: magenta
Proof-Auditor — Independent Rubric Verdict
Behavioral Boundary
You AUDIT and SCORE — you do not modify the audited deliverable, do not edit the rubric, and do not in
Alternatives in Development
- Context Driven Development — Guide for implementing and maintaining context as a managed artifact alongside code, enabling consis 31.9k ★
- Metadata Extractor — Extracts paper metadata (authors, date, venue, fields, DOI/arxiv ID) and a paper-quality assessment (credibili 95 ★
- Patent Researcher — Expert in patent prior art searching, patentability assessment, and competitive intelligence 91 ★
Full documentation available on GitHub
View Source RepositoryRelated Agents
Risk Scorer
Fast risk assessment of PR changes. Computes a 0-100 risk score from 6 weighted factors and recommends which r
Opus5 Prompt Expert
Expert on prompting Claude Opus 5 and on how Claude Code loads instructions. Use before adding, rewording, or
Demo Agent Objection Panel
Runs three named stakeholders against a recommendation in a separate context and returns one ranked summary, n
Competitive Strategist
Maps everyone a new product would have to beat, from direct competitors to substitutes to doing nothing. Recor
War Room
Stress tests a product concept by playing each stakeholder in turn (engineering lead, PM, sales, support, skep
Disagreement Judge
Adjudicate disputed checks when two claim-verifiers disagree. Sees the artifact, the disputed checks only, and
Related Skills
Muse:Debate
Run a 3-round adversarial debate between two personas on a question. Round 1 is opening statements, Round 2 is
QA Lab
Two-agent QA and application security assessment workspace. Two AI coding agents from two providers run indepe
Forgeflow Adoption
Print a concise Forgeflow adoption pack with fit guidance, first-trial steps, proof boundaries, and decision r