What problem does it solve? Long agent sessions degrade: analysis, implementation, and review blur together in one context, causing self-review blindness, bundled fixes, and the fix-break-fix spiral. This Skill structures software work into phased workflows with fresh contexts and independent sub-agent verification. ## Core Features & Use Cases - Seven End-to-End Workflows: Bug fix, feature implementation, code review, code quality optimization, architecture upgrade, SRS coverage analysis, and quality reporting, all following an Understand → Plan → Execute → Review → Report pattern. - Sub-Agent Isolation Rules: Reviewer never equals implementer, fixer never equals reviewer, and auto-fix retries are capped at 2 cycles before escalation. - Context Hygiene Discipline: Explicit rules for when to reset context (/new, /compress, delegate_task) plus a Rule of Three escape hatch when repeated fixes fail. - Use Case: A developer reports a bug; the agent first reproduces it with a failing test, traces the root cause, applies a minimal fix, then delegates an independent sub-agent review and verifies every claim by re-running tests before reporting. ## Quick Start Ask the agent to fix a bug using the structured workflow: reproduce it with a failing test first, analyze the root cause, apply a minimal fix, then run an independent sub-agent review.
