What problem does it solve? Engineering requests arrive with wildly different scope — a one-line typo fix and a multi-week epic should not follow the same process. This Skill acts as a single entry point that classifies any request into a lane (quick-fix, task, epic, portfolio, research, or refactoring), proposes the right skill chain, and waits for confirmation before any work begins, preventing both under-process on big work and over-process on small fixes. ## Core Features & Use Cases - Lane classification with ambiguity rules: Signal tables classify requests by magnitude, with one targeted clarifying question when signals conflict (e.g., fix-vs-refactor, task-vs-epic). - Confirmed execution plans: Presents the detected lane, reasoning, proposed skill chain, and skipped skills with reasons — no skill fires until the user approves. - Lane-specific workflows: Quick-fix runs debugging plus TDD; task runs comprehension, spec gate, planning, worktrees, and review; epic adds business intake, HLD, decomposition, deployment, and production-readiness gates; portfolio sequences epics via WSJF and WIP limits. - Cost observability and circuit breaker: The bundled cost-checkpoint script brackets every step and subagent dispatch with token snapshots, appends rows to a cross-repo JSONL ledger, and can halt execution when a session token cap is breached. - Use Case: A user types "/engineer the login endpoint returns 500 after password reset". The Skill classifies it as quick-fix, proposes systematic-debugging plus TDD plus commit-discipline, waits for confirmation, then executes the chain with per-step cost tracking. ## Quick Start Ask the AI to run /engineer followed by a description of your engineering task, then confirm the proposed lane and skill chain it presents.
