Doubt-Driven Review

Adversarial pre-mortem gate for risky diffs, migrations, and 'this is safe' claims

di pimenov · pimenov/codex-first-skills-pack

Promosso ★ 9.2/10

Doubt-Driven Review — Adversarial pre-mortem gate for risky diffs, migrations, and 'this is safe' claims

Cosa fa

A single self-contained methodology skill: state the claim, extract artifact+contract, run an adversarial 'find what's wrong' review (fresh subagent, external CLI with approval, or labeled degraded self-review), classify every finding, and stop only on a defined exit condition. Triggers before production, auth, data, billing, or irreversible changes are allowed to stand as 'safe' or 'done'.

Rapporto di test

Ran a real prod-column-drop scenario baseline-vs-skill: baseline gave a decent ad-hoc safety checklist, the skill forced an explicit CLAIM/CONTRACT/adversarial-findings structure that surfaced a missing-backup gap and a lock-contention-under-load edge case the baseline only hinted at, and ended in an auditable 'revise, don't approve yet' decision instead of a soft 'looks fine if...'.

Testato il: 2026-07-14 · Claude Code 2.x (agent harness)

Installazione

git clone https://github.com/pimenov/codex-first-skills-pack
cd codex-first-skills-pack
mkdir -p ~/.claude/skills
cp -r skills/doubt-driven-review ~/.claude/skills/doubt-driven-review

Comandi e prompt di esempio

/doubt-driven-reviewAdversarial pre-mortem gate for risky diffs, migrations, and 'this is safe' claims

Gli skill si attivano con richieste in linguaggio naturale, senza comandi da ricordare. Dopo l'installazione, prompt come questi lo attivano (in inglese):

Is it actually safe to drop this column in production?
Pre-mortem this migration before I ship it tomorrow
Challenge my claim that this refactor is behaviour-preserving

Rapporto completo →