domains / families / software-engineering / audit-choices

audit-choices

separate

Audit the choices an implementing agent made, not its diff — a pure decision audit that traces the session's history into a choices ledger, changes no code, and never blocks an unsupervised run. Working code still embeds architecture the user never chose; surface it because future work inherits it. Use when the user wants to review the decisions the AI made on their behalf, before merging or committing AI-implemented work, when integrating a delegated subagent's pass, or when a fix "works" but might be a point fix.

When to use

• reviewing decisions the AI made on the user's behalf
• auditing work before merging or committing
• integrating a delegated subagent's pass
• evaluating fixes that work but might be point fixes
• tracing session history and subagent reports
• surfacing architectural decisions not in the original spec

When not to use

• modifying code
• blocking an unsupervised run
• inspecting execution diffs for bugs
• acting on verdicts or redoing unsound choices

Skill page

Source