assumption-auditlisted

# Assumption Audit ## Purpose Make unverified claims visible before they become implementation decisions. ## Preserves Evidence-based judgment. ## Required Evidence - Proposed plan, answer, or design. - Claims the plan depends on. - Available code, docs, logs, tests, metrics, or constraints. ## Failure Signs - Confidence is treated as proof. - The riskiest assumption is buried. - Locally checkable claims are sent back to the user instead of verified. ## When To Use - A plan depends on behavior that has not been checked. - The user or assistant says "obviously", "just", "safe", or "simple". - A design depends on API, framework, runtime, user, or operational assumptions. - AI produced a fluent answer without evidence. ## When Not To Use - Requirements already verified by tests, docs, or code references. - Cosmetic changes with no behavioral or operational effect. - Early brainstorming where evaluation is explicitly deferred. ## Inputs Expected - Proposed plan, answer, or design. - Any available code, docs, logs, tests, metrics, or constraints. - The decision that depends on the assumptions. ## Output Expected ```md Verified: Likely but unproven: Risky: Unknown: Recommended next check: ``` ## Process 1. Extract concrete assumptions from the plan. 2. Classify each as factual, technical, product, operational, or social. 3. Check locally verifiable assumptions in code or docs before asking the user. 4. Mark each assumption as verified, likely, risky, or unknown. 5