What problem does it solve? AI agents often claim they followed mandatory workflow constraints (write tests first, get plan approval, run verification) without actually doing so. This Skill independently audits the conversation trail and tool call ordering to verify whether HARD-GATEs and constraints were genuinely followed, not just claimed. ## Core Features & Use Cases - HARD-GATE Auditing: Extracts mandatory constraints from each invoked skill's SKILL.md and checks them against actual tool call evidence. - Violation Classification: Distinguishes BLOCK-level HARD-GATE violations from WARN-level constraint violations and INFO-level best-practice skips. - Compliance Reporting: Produces a per-skill compliance summary table with specific evidence and remediation steps for each violation. - Use Case: After a cook workflow finishes, run a discipline audit to confirm tests were written before implementation code, plan approval was obtained, and verification output was actually captured. ## Quick Start Audit the current session for constraint compliance by asking the agent to run a constraint check on all skills invoked in this workflow.
