AI Work Product Policy

Purpose

AI-generated work must be reviewable, attributable to a prompt/task, and safe to integrate.

Required For Non-trivial AI Work

• Source prompt or task graph summary.
• Allowed paths and forbidden paths.
• Changed files list.
• Validation commands.
• Risks and follow-ups.
• Human reviewer.

Review Focus

• Hidden broad refactors.
• Incorrect claims about implemented features.
• Stale docs.
• Secret leakage.
• License or copied-content risk.
• Security-sensitive behavior added without approval.

Forbidden

• Raw private prompt dumps in repo.
• AI-generated legal, tax, immigration, or investment commitments.
• Production secrets or credentials.
• Undocumented backend, billing, auth, or deployment changes.