qa-false-positive-detection

Run the false-positive-check command with the story ID and claimed fix description. It performs a remove-and-reapply causation test, checks assumptions and bias factors, then outputs a confidence score with a verification report.

A false positive is a fix that appears to resolve a bug but does not address the root cause, such as placebo fixes, timing coincidences, environment-dependent behavior, or incomplete edge case handling. These pass superficial review but regress later.

The score weights seven factors: remove test (0.25), assumptions verified (0.20), old code fails (0.15), new code succeeds (0.15), not self-healing (0.10), negative cases (0.10), and mechanism explained (0.05). Scores of 0.85 or higher indicate high confidence.

It runs during the review-build phase whenever a story is identified as a bug fix. The confidence score is included in the QA report, and scores below 0.65 strongly weigh toward rejecting the fix.

Red flags include testing only the happy path, inability to reproduce the original bug, unclear fix mechanism, timing or environment dependence, missing regression tests, undocumented assumptions, and considering only one explanation for the bug.