discipline

What problem does it solve? AI agents repeatedly fail in predictable ways: fabricating URLs and commands, ignoring user constraints, claiming work is done without verifying it, retrying blindly, and drifting out of scope. This Skill loads an operating protocol of 15 rules distilled from 120+ real sessions to prevent those failure modes before they happen. ## Core Features & Use Cases - Failure-Prevention Rules: 15 concrete rules covering fabrication, constraint tracking, verification, investigation, scoping, retry discipline, evidence-based claims, and writing style. - Pre-Flight and Post-Flight Checklists: Mental checklists to run before every action and before declaring work done, including git status checks and output verification. - Session Failure Escalation: Named failure states (fabrication spiral, retry loop, constraint drift, scope creep, blame mode) with explicit stop conditions. - Use Case: Load the protocol at the start of any coding or investigation task so the agent verifies claims with evidence, respects stated technology choices, commits changes, and never invents commands or file contents. ## Quick Start Ask your agent to load the discipline protocol before starting the next task so it follows the 15 failure-prevention rules throughout the session.