What problem does it solve? When you step away from a coding session, a default assistant stalls on every ambiguity, asking questions nobody is there to answer. This Skill flips the default stance from "ask if unsure" to "act if unsure, log what you chose", so research and long-running tasks keep moving while you are away. ## Core Features & Use Cases - Autonomous operating mode: Picks the less-destructive option when ambiguous, records the decision in a monthly dev log, and keeps working instead of pausing. - Guardrails and time-boxing: Stops for credentials, destructive operations on shared state, paid resources, scope explosions, and repeated failures; enforces budgets like ~2 hours per bug and ~1 day per architectural experiment. - Review and notification loop: Triggers /review-code and /review-docs after changes (max two cycles), commits in small logical units, and sends push notifications via ntfy.sh at semantic milestones. - Use Case: You kick off a multi-hour refactoring before leaving for the day. The agent plans first, works through the task list, writes tests, commits incrementally, logs decisions to docs/dev-log/YYYY-MM.md, and pings your phone only when it finishes or genuinely needs you. ## Quick Start Type /unattended to switch the current session into autonomous mode for your long-running task.
