doubt-driven-development

Release readiness validation requires converting delivery claims into falsifiable statements and challenging them with adversarial evidence like logs and code graphs. This Skill runs a bounded skeptical review to ensure production claims are defensible before release.

Adversarial review is a bounded process that challenges post-implementation behavioral, security, and workflow claims using extracted evidence references. It runs up to three structured cycles of doubt, reconciliation, and targeted verification to produce a final verdict.

Security validation runs by applying an adversarial loop against your delivery claim using memory, code search, and verification outputs. It escalates blockers for authorization gaps, receipt gaps, and security risks, ultimately recording residual risk in a doubt-review-report.

Yes, doubt review applies to post-implementation behavioral, security, workflow, and receipt claims. It extracts structured evidence references and runs targeted verification to ensure claims with thin or indirectly inferred evidence are thoroughly scrutinized.

When evidence is thin or missing, the adversarial loop escalates blockers for missing proof, contract breaks, and unhandled edge states. It outputs a doubt-review-report with status, confidence, nextAction, blockers, and records residual risk when allowed.

The adversarial review loop runs up to three structured cycles of doubt, reconciliation, and targeted verification, ending with an explicit STOP and verdict. It prevents rubber-stamping by escalating blockers and outputting a final doubt-review-report.