aif-debuggerlisted

# aif-debugger ## Trigger Use when an observed failure and expected behavior are known or can be made explicit. Do not trigger for a hypothetical concern with no observable signal. ## Inputs - observed output, expected behavior, and available reproduction evidence - Any explicitly relevant canonical policy or workflow. ## Procedure 1. Read the relevant project context, specifications, ADRs, and recent evidence. 2. Build the tightest safe pass/fail loop that exercises the reported symptom. If no valid loop is possible, state the missing artifact or access instead of guessing. 3. Reproduce and minimize the failure. For a hard bug, rank three to five falsifiable hypotheses before instrumenting one variable at a time. 4. Identify the root cause with evidence. Modify code only when the user asked for a fix; otherwise return the diagnosis and safe next action. 5. For an authorized fix, add a regression test at the real behavior seam, verify red then green when safely expressible, and re-run the original loop. 6. Remove temporary instrumentation and record residual risk. ## Exact outputs Return reproduction, minimized case, tested hypotheses, root cause, evidence, regression status, residual risk, and safe next action. ## Stop conditions Stop when the diagnosis or authorized fix is verified, when no honest feedback loop can be built, or when a decision outside the request is needed. State the blocker rather than guessing.