bugfix-quick

> **AI-consumed reference.** Optimized for Claude to read during execution. > Human-readable explanation: see [docs/architecture/HIERARCHICAL_PLANNING.md](../../../docs/architecture/HIERARCHICAL_PLANNING.md) > or [docs/getting-started/](../../../docs/getting-started/) depending on topic. # Quick Bugfix For bugs only. Features/refactors → run-orchestrator. --- ## Core Principle **NO FIXES WITHOUT ROOT CAUSE.** Fix at SOURCE, not symptom. ## Process (4 Steps) | Step | Name | Agent | Statusline | |---|---|---|---| | S1 | Investigate | `lead` (inline) — uses Read/Grep/Bash to reproduce | `bugfix S1` | | S2 | Test RED | `tester` — writes failing test | `bugfix S2` | | S3 | Fix GREEN | `architect` / `frontend` / `mobile` (per artifact, via `agent-detector`) | `bugfix S3` | | S4 | Verify | `tester` — runs full suite | `bugfix S4` | **Announce per step.** When transitioning, surface to the user: `─── Step S{N} · {Name} ─── Dispatching {agent}…`. Also update `run-state.json#current_step` (one of: `investigate`, `test-red`, `fix-green`, `verify`) and `#active_agent` so the statusline reflects reality. ### 1. Investigate - Read error + stack trace - Reproduce the bug - `git log` / `git diff` for recent changes - Trace backward: where did bad data originate? - **Context economy** (`rules/core/context-economy.md`) — Grep for the failing symbol first, then Read with `offset`+`limit` on the matched lines only. Don't read the full file unless the bug spans it. ### 2. Write Failing...