Repo Truth Auditor

Overview

This agent is designed to act as a 'Pressure Test Mode' auditor, forcing any AI model (like Grok) to anchor its critique strictly to a predefined set of canonical project truths. It prevents hallucination by establishing a hard source of truth regarding repository status, financial models, and live infrastructure.

Capabilities

• Truth Anchoring: Forces the LLM to operate only within specified facts (e.g., Square as the payment rail, specific URLs).
• Risk Identification: Systematically calls out security vulnerabilities, trust issues, and high launch risks in any provided text.
• Assumption Testing: Explicitly separates proven facts from mere inferences or outdated 'mission mythology.'
• Structured Output: Guarantees a highly structured response format including BASELINE ACK, TOP RISKS, WEAKEST ASSUMPTIONS, and NEXT TESTS.

Example Use Cases

1. Pre-Launch Review: Before submitting marketing copy or feature documentation, run it through this agent to ensure no outdated financial models (like old DAO references) slip through.
2. Code Audit Simulation: When reviewing a pull request that touches sensitive areas, use this prompt to challenge any assumptions about live API endpoints or control mechanisms.
3. Stakeholder Alignment: Use it when multiple documents exist for a project; the agent forces consensus by prioritizing the 'canonical workspace' truth over conflicting narratives.