brewcode:plan

What this skill does

Create Plan — [task-dir or SPEC path or plan file]

<instructions>

## /brewcode:plan Instructions

**ROLE:** Plan Creator | **OUTPUT:** PLAN.md + phases/*.md + KNOWLEDGE.jsonl + artifacts/ + backup/

### Input Detection

| Input | Action |
|-------|--------|
| Path to `{TS}_{NAME}_task/` dir | Read SPEC.md from it |
| Path to `SPEC.md` file | Derive task dir from parent |
| `.claude/plans/LATEST.md` or plan file | Plan Mode: parse plan, create task dir, skip SPEC |
| Empty | Check `.claude/TASK.md` quick ref for latest task dir |

### Flag Parsing

Parse `$ARGUMENTS` for flags BEFORE input detection:

| Flag | Effect |
|------|--------|
| `-n`, `--noask` | Skip all user questions, auto-approve defaults |

Strip flag from `$ARGUMENTS`. Remaining text = path.

### Workflow (SPEC input)

0. **Check Adapted Templates** (REQUIRED FIRST)

   **EXECUTE** using Bash tool:
   ```bash
   TMPL_DIR=".claude/tasks/templates"
   PLUGIN_TMPL="$BC_PLUGIN_ROOT/skills/setup/templates"
   echo "--- Project templates ---"
   test -f "$TMPL_DIR/PLAN.md.template" && echo "PLAN.md.template OK" || echo "PLAN.md.template MISSING"
   test -f "$TMPL_DIR/phase.md.template" && echo "phase.md.template OK" || echo "phase.md.template MISSING"
   test -f "$TMPL_DIR/phase-verify.md.template" && echo "phase-verify.md.template OK" || echo "phase-verify.md.template MISSING"
   test -f "$TMPL_DIR/phase-final-review.md.template" && echo "phase-final-review.md.template OK" || echo "phase-final-review.md.template MISSING"
   test -f "$TMPL_DIR/phase-fix.md.template" && echo "phase-fix.md.template OK" || echo "phase-fix.md.template MISSING"
   echo "--- Plugin fallback templates ---"
   test -f "$PLUGIN_TMPL/PLAN.md.template" && echo "PLAN.md.template FALLBACK OK" || echo "PLAN.md.template FALLBACK MISSING"
   test -f "$PLUGIN_TMPL/phase.md.template" && echo "phase.md.template FALLBACK OK" || echo "phase.md.template FALLBACK MISSING"
   test -f "$PLUGIN_TMPL/phase-verify.md.template" && echo "phase-verify.md.template FALLBACK OK" || echo "phase-verify.md.template FALLBACK MISSING"
   test -f "$PLUGIN_TMPL/phase-final-review.md.template" && echo "phase-final-review.md.template FALLBACK OK" || echo "phase-final-review.md.template FALLBACK MISSING"
   test -f "$PLUGIN_TMPL/phase-fix.md.template" && echo "phase-fix.md.template FALLBACK OK" || echo "phase-fix.md.template FALLBACK MISSING"
   ```

   **Template resolution order:**
   1. Project templates: `.claude/tasks/templates/{name}.template` (adapted by `/brewcode:setup`)
   2. Plugin fallback: `$BC_PLUGIN_ROOT/skills/setup/templates/{name}.template`

   > **STOP if BOTH locations MISSING for any template** -- Run `/brewcode:setup` first to get v3 templates.
   > If project templates missing but plugin fallback exists: WARN "Re-run /brewcode:setup for v3 project-adapted templates. Using plugin defaults."

1. **Read SPEC**

   - Resolve input path per Input Detection table
   - Read SPEC.md from task directory
   - Extract: goal, requirements, analysis, context files, risks, decisions

2. **Scan Project for Reference Examples**

   - Find 1-2 canonical files per expected phase type (controller, service, test, etc.)
   - These become Reference Examples (R1, R2...) in PLAN.md

### Dynamic Agent Resolution

Before assigning agents to phases, check for project team agents:

1. If `.claude/teams/` exists — read `team.md` for agent roster with domains
2. If `.claude/agents/` has project agents — list available
3. Match agent domain to phase task area
4. Priority: **team agent > project agent > plugin agent > system agent**
5. If agent refuses (Task Acceptance Protocol) — re-delegate to suggested colleague (max 2 retries)

> Always fall back to plugin agents when no project agents match the task domain.

3. **Generate Phase Breakdown** (5-12 phases)

   Based on SPEC analysis and project structure:
   - Each phase = one logical unit of work
   - Dependencies between phases identified
   - Agents assigned per phase
   - Verification phases (NV) after each execution phase
   - Final Review (FR) as last phase

4. **Present Phases to User** (AskUserQuestion)

   **If `--noask`:** Skip. Auto-approve all phases.

   **Otherwise:** Use AskUserQuestion to present the proposed phase split:
   - Phase count and descriptions
   - Agent assignments
   - Dependency chain
   - User can approve, adjust, or request changes

5. **Generate Artifacts**

   Read templates per resolution order from Step 0 (project first, plugin fallback second).

   **5.1 Create directory structure**

   ```
   .claude/tasks/{TS}_{NAME}_task/
   ├── phases/           <-- NEW: individual phase files
   ├── artifacts/
   ├── backup/
   ├── KNOWLEDGE.jsonl   (0-byte empty file)
   └── PLAN.md           (slim, with Phase Registry)
   ```

   **EXECUTE** using Bash tool:
   ```bash
   TASK_DIR=".claude/tasks/{TS}_{NAME}_task"
   mkdir -p "$TASK_DIR/phases" "$TASK_DIR/artifacts" "$TASK_DIR/backup"
   touch "$TASK_DIR/KNOWLEDGE.jsonl"
   ```

   **5.2 Generate phase files** (from templates)

   For EACH execution phase (1, 2, 3, ...):
   - Read `phase.md.template`
   - Fill placeholders with SPEC-derived content:
     - `{PHASE_NUM}` -- phase number
     - `{PHASE_NAME}` -- descriptive name (kebab-case for filename, Title Case for heading)
     - `{AGENT}` -- assigned agent
     - `{AGENT_ROLE}` -- one-line agent role description
     - `{OBJECTIVE}` -- concrete objective from SPEC analysis
     - `{CONTEXT_FILES}` -- table rows: files the agent needs to read/modify
     - `{REFERENCES}` -- table rows: reference examples, docs, existing patterns
     - `{TASK_LIST}` -- numbered task list with specific, actionable items
     - `{CONSTRAINTS}` -- project-specific constraints (from SPEC + project rules)
     - `{EXIT_CRITERIA}` -- measurable exit criteria (builds, tests pass, lint clean, etc.)
     - `{ARTIFACT_DIR}` -- e.g., `1-1e` (Phase 1 Execution, iter 1)
     - `{ADDITIONAL_ARTIFACTS}` -- files created/modified by this phase
   - Write to: `phases/{N}-{name}.md` (e.g., `phases/1-create-entity.md`)

   For EACH verification phase:
   - Read `phase-verify.md.template`
   - Fill placeholders:
     - `{PHASE_NUM}` -- matches the execution phase being verified
     - `{PHASE_NAME}` -- same name as the execution phase
     - `{VERIFY_AGENT}` -- tester or reviewer
     - `{FILES_TO_REVIEW}` -- files created/modified by the execution phase
     - `{VERIFICATION_CHECKLIST}` -- checklist items derived from exit criteria of execution phase
     - `{AGAINST_REFERENCES}` -- reference examples to compare against
     - `{ARTIFACT_DIR}` -- e.g., `1-1v` (Phase 1 Verification, iter 1)
     - `{AGENT}` -- same as `{VERIFY_AGENT}`
   - Write to: `phases/{N}V-verify-{name}.md`

   For Final Review:
   - Read `phase-final-review.md.template`
   - Fill placeholders:
     - `{FR_AGENTS}` -- list of review agents (typically reviewer + tester + architect)
     - `{COMPLETION_CRITERIA}` -- from SPEC.md goals/decisions, copied to PLAN.md Completion Criteria
     - `{REVIEW_CHECKLIST}` -- comprehensive checklist covering all phases
     - `{FILES_CHANGED}` -- aggregate of all files created/modified across all phases
     - `{ARTIFACT_DIR}` -- `FR-1e` (Final Review, iter 1)
   - Write to: `phases/FR-final-review.md`

   **5.3 Generate PLAN.md** (slim v3 format)

   Using `PLAN.md.template` (project-adapted or plugin fallback):
   - Fill Phase Registry table with ALL generated phase files
   - Each row references the corresponding `phases/{file}.md`
   - Completion Criteria from SPEC.md decisions/goals
   - Agents table from project analysis
   - If `.claude/teams/` exists: populate `### Project Agents` table from team.md roster
   - Technology Choices from SPEC analysis
   - Role Constraints from project rules

   **5.4 Technology Choices**

   For each non-trivial choice (library, pattern, approach):
   - Document in PLAN.md under Technology Choices section
   - Include rationale + alternatives considered and rejec