What problem does it solve? Active work sessions often degrade into streams of agent process narration, vague error reports, and unnecessary questions that push agent-owned work back onto the user. This Skill enforces a discipline where every user-facing message leads with actionable state, keeps agent-executable work agent-owned, and reports errors with evidence and a recovery path. ## Core Features & Use Cases - Actionable message structure: Leads each turn with completed results, current blockers, or decisions needed, using a Done/Now/Next state shape instead of process logs. - Bounded user actions and questions: Decomposes required user steps into concrete ordered actions and asks only consequential questions that project evidence cannot resolve. - Evidence-based error and completion reporting: Reports failures with symptom, evidence, cause, recovery, and verification criteria, and ties completion claims to test, build, or read-back evidence. - Use Case: During a multi-turn debugging session, the agent reports a failing validation gate with the exact error output and confirmed cause, applies the fix itself, and reports the green test run—rather than narrating each tool call or asking the user to run commands. ## Quick Start Apply the interaction-discipline skill to keep all progress updates, blocker reports, and completion messages in this session focused on verified state and concrete next actions.
