maxsimcli 5.0.7 → 5.1.0

package/dist/assets/templates/workflows/execute-plan.md

@@ -1,1090 +1,300 @@
 <purpose>
-Execute a phase plan (loaded from GitHub issue comment or local PLAN.md) and post the outcome summary as a GitHub comment. Per-task board transitions keep the project board current throughout execution.
+Template prompt for individual plan execution by a spawned executor agent. This file defines what each executor agent receives and must produce. It is self-contained — the executor has no access to the orchestrator's context and must work from what is explicitly provided here.
+
+Executor agents run in isolated git worktrees. Each agent executes all tasks in one plan, commits atomically per task, updates the GitHub board, and posts a summary comment on the phase issue.
 </purpose>
 
-<required_reading>
-Read STATE.md before any operation to load project context.
-Read config.json for planning behavior settings.
+<agent_type>executor</agent_type>
+<isolation>worktree</isolation>
 
-@~/.claude/maxsim/references/git-integration.md
+<required_reading>
+Read these files immediately upon start, before any other action:
+- .planning/STATE.md — project state and position
+- .planning/config.json — planning behavior settings (if exists)
+- ./CLAUDE.md — project-specific coding conventions (if exists)
 </required_reading>
 
-MAXSIM provides CLI commands (`github create-phase`, `github list-phases`, etc.) for structured operations.
-
 <process>
 
-<step name="init_context" priority="first">
-Load execution context (paths only to minimize orchestrator context):
-
-```bash
-INIT=$(node ~/.claude/maxsim/bin/maxsim-tools.cjs init execute-phase "${PHASE}")
-```
-
-Extract from init JSON: `executor_model`, `commit_docs`, `phase_dir`, `phase_number`, `plans`, `summaries`, `incomplete_plans`, `state_path`, `config_path`, `phase_issue_number`, `task_mappings`.
-
-If `.planning/` missing: error.
-</step>
+## Step 1 — Load Context
 
-<step name="load_plan_from_github">
-
-## Plan Loading -- GitHub First
-
-When the orchestrator passes `github_context` (phase_issue_number and plan_comment_body), use it directly:
-
-1. The plan content is in `plan_comment_body` passed from the orchestrator. Parse it in memory -- do NOT read a local PLAN.md file.
-2. Extract frontmatter fields: `wave`, `autonomous`, `objective`, `requirements`, `task_mappings`, `gap_closure`.
-3. Extract the task list and verification criteria from the comment body.
-
-**External edit detection (WIRE-06):** Before beginning execution, check if the plan comment was modified since the orchestrator read it:
+Load execution context:
 
 ```bash
-node ~/.claude/maxsim/bin/maxsim-tools.cjs github detect-external-edits --phase-number "$PHASE_NUMBER"
+INIT=$(node ~/.claude/maxsim/bin/maxsim-tools.cjs init execute-phase "${PHASE_NUMBER}")
 ```
 
-If external edits detected: warn user and offer to re-read the plan before proceeding.
+Extract: `phase_dir`, `phase_number`, `phase_name`, `executor_model`, `commit_docs`, `phase_issue_number`.
 
-**If no GitHub context is available:**
-
-GitHub Issues is the source of truth for plans. If no `phase_issue_number` or `plan_comment_body` was passed, report the error and exit:
-
-```
-Plan content must be provided via GitHub Issue comment. Ensure GitHub integration is configured.
-```
-</step>
+If `.planning/` is missing: exit immediately with `RESULT: FAIL — .planning/ directory not found`.
 
-<step name="record_start_time">
+Record start time:
 ```bash
 PLAN_START_TIME=$(date -u +"%Y-%m-%dT%H:%M:%SZ")
 PLAN_START_EPOCH=$(date +%s)
 ```
-</step>
 
-<step name="parse_segments">
+## Step 2 — Parse Plan Content
 
-Identify checkpoint type from plan content:
+The plan content is provided in the orchestrator's prompt via `<plan_content>` and `<github_context>` blocks.
 
-```bash
-# If plan loaded from GitHub comment, scan comment body for checkpoint markers
-# If plan loaded from local file:
-grep -n "type=\"checkpoint" .planning/phases/XX-name/{phase}-{plan}-PLAN.md
-```
-
-**Routing by checkpoint type:**
-
-| Checkpoints | Pattern | Execution |
-|-------------|---------|-----------|
-| None | A (autonomous) | Single subagent: full plan + post SUMMARY comment + commit |
-| Verify-only | B (segmented) | Segments between checkpoints. After none/human-verify → SUBAGENT. After decision/human-action → MAIN |
-| Decision | C (main) | Execute entirely in main context |
+Parse from the plan content:
+- **Plan ID** — e.g. `03-02`
+- **Objective** — one-line description of what this plan builds
+- **Wave** — wave number (for ordering context only)
+- **Task list** — ordered list of tasks, each with: name, description, files to modify, acceptance criteria, done criteria
+- **Success criteria** — overall plan success criteria
+- **Task sub-issue numbers** — from `task_mappings` in github_context
 
-**Pattern A:** init_agent_tracking → spawn Task(subagent_type="executor", model=executor_model) with prompt: execute plan (content from GitHub comment or local path), autonomous, all tasks + post SUMMARY comment + commit, follow deviation/auth rules, move task sub-issues on board as each task starts/completes, report: plan name, tasks, summary comment URL, commit hash → track agent_id → wait → update tracking → report.
+Detect checkpoint type from plan content:
+- No checkpoints → **Pattern A** (autonomous, single pass)
+- Checkpoint markers present → **Pattern B** (segmented execution)
 
-**Pattern B:** Execute segment-by-segment. Autonomous segments: spawn subagent for assigned tasks only (no SUMMARY/commit). Checkpoints: main context. After all segments: aggregate, post SUMMARY as GitHub comment, commit. See segment_execution.
+## Step 3 — Execute Tasks (Pattern A — Autonomous)
 
-**Pattern C:** Execute in main using standard flow (step name="execute").
+Execute all tasks in sequence. For each task:
 
-Fresh context per subagent preserves peak quality. Main context stays lean.
-</step>
+### 3a — Announce task start
 
-<step name="init_agent_tracking">
+Move task sub-issue to "In Progress" on the board:
 ```bash
-if [ ! -f .planning/agent-history.json ]; then
-  echo '{"version":"1.0","max_entries":50,"entries":[]}' > .planning/agent-history.json
-fi
-rm -f .planning/current-agent-id.txt
-if [ -f .planning/current-agent-id.txt ]; then
-  INTERRUPTED_ID=$(cat .planning/current-agent-id.txt)
-  echo "Found interrupted agent: $INTERRUPTED_ID"
-fi
-```
-
-If interrupted: ask user to resume (Task `resume` parameter) or start fresh.
-
-**Tracking protocol:** On spawn: write agent_id to `current-agent-id.txt`, append to agent-history.json: `{"agent_id":"[id]","task_description":"[desc]","phase":"[phase]","plan":"[plan]","segment":[num|null],"timestamp":"[ISO]","status":"spawned","completion_timestamp":null}`. On completion: status → "completed", set completion_timestamp, delete current-agent-id.txt. Prune: if entries > max_entries, remove oldest "completed" (never "spawned").
-
-Run for Pattern A/B before spawning. Pattern C: skip.
-</step>
-
-<step name="segment_execution">
-Pattern B only (verify-only checkpoints). Skip for A/C.
-
-1. Parse segment map: checkpoint locations and types
-2. Per segment:
-   - Subagent route: spawn executor for assigned tasks only. Prompt: task range, plan content (from GitHub comment or local path), read full plan for context, execute assigned tasks, track deviations, move task sub-issues to "In Progress" when started and "Done" when completed, NO SUMMARY/commit. Track via agent protocol.
-   - Main route: execute tasks using standard flow (step name="execute")
-3. After ALL segments: aggregate files/deviations/decisions → post SUMMARY as GitHub comment (`github post-comment` with type=summary) → commit → self-check:
-   - Verify key-files.created exist on disk with `[ -f ]`
-   - Check `git log --oneline --all --grep="{phase}-{plan}"` returns ≥1 commit
-   - Check for `## Self-Check: PASSED` or `## Self-Check: FAILED` and append to summary comment body
-
-   **Known Claude Code bug (classifyHandoffIfNeeded):** If any segment agent reports "failed" with `classifyHandoffIfNeeded is not defined`, this is a Claude Code runtime bug — not a real failure. Run spot-checks; if they pass, treat as successful.
-</step>
-
-<step name="load_prompt">
-
-The plan content IS the execution instructions. Follow exactly. If plan references CONTEXT.md: honor user's vision throughout.
-
-The plan content is already in memory from the load_plan_from_github step (loaded from the GitHub Issue comment passed by the orchestrator).
-</step>
-
-<step name="pre_execution_gates">
-Validate requirements before execution. These gates ensure spec-driven development.
-
-**Gate G1: Requirement Existence** — All requirement IDs from the plan's frontmatter must exist in REQUIREMENTS.md.
-
-```bash
-# Extract requirement IDs from plan frontmatter (in memory if loaded from GitHub, or from file)
-REQ_IDS=$(grep "^requirements:" "$PLAN_PATH" | sed 's/requirements:\s*\[//;s/\]//;s/,/ /g;s/"//g;s/'\''//g' | tr -s ' ')
-if [ -n "$REQ_IDS" ]; then
-  G1_RESULT=$(node ~/.claude/maxsim/bin/maxsim-tools.cjs verify requirement-existence $REQ_IDS)
-  G1_VALID=$(echo "$G1_RESULT" | node -e "process.stdin.on('data',d=>console.log(JSON.parse(d).valid))")
-  if [ "$G1_VALID" != "true" ]; then
-    G1_MISSING=$(echo "$G1_RESULT" | node -e "process.stdin.on('data',d=>console.log(JSON.parse(d).missing.join(', ')))")
-    echo "GATE G1 FAILED: Requirements not found in REQUIREMENTS.md: $G1_MISSING"
-    echo "Fix: Add missing requirements to REQUIREMENTS.md or update plan frontmatter."
-    # STOP execution — this is a hard gate
-  fi
-fi
-```
-
-**Gate G2: Requirement Status** — Requirement IDs must not already be marked Complete.
-
-```bash
-if [ -n "$REQ_IDS" ]; then
-  G2_RESULT=$(node ~/.claude/maxsim/bin/maxsim-tools.cjs verify requirement-status $REQ_IDS)
-  G2_VALID=$(echo "$G2_RESULT" | node -e "process.stdin.on('data',d=>console.log(JSON.parse(d).valid))")
-  if [ "$G2_VALID" != "true" ]; then
-    G2_COMPLETE=$(echo "$G2_RESULT" | node -e "process.stdin.on('data',d=>console.log(JSON.parse(d).already_complete.join(', ')))")
-    echo "GATE G2 WARNING: Requirements already marked complete: $G2_COMPLETE"
-    echo "This plan may be re-implementing completed work. Proceeding with warning."
-    # WARNING only — log but continue (re-execution scenarios are valid)
-  fi
-fi
-```
-</step>
-
-<step name="previous_phase_check">
-```bash
-node ~/.claude/maxsim/bin/maxsim-tools.cjs phases list --type summaries --raw
-# Extract the second-to-last summary from the JSON result
-```
-If previous phase has unresolved issues (check summary comments on phase issue for "Issues Encountered" / "Next Phase Readiness" blockers): AskUserQuestion(header="Previous Issues", options: "Proceed anyway" | "Address first" | "Review previous").
-</step>
-
-<step name="execute">
-Deviations are normal — handle via rules below.
-
-1. Read @context files from prompt
-2. Per task:
-
-   **Before starting each task (WIRE-04 board transition):**
-   - Look up the task's sub-issue number from `task_mappings` in the GitHub context.
-   - Move the task sub-issue to "In Progress":
-     ```bash
-     node ~/.claude/maxsim/bin/maxsim-tools.cjs github move-issue --issue-number $TASK_SUB_ISSUE_NUMBER --status "In Progress"
-     ```
-
-   - `type="auto"`: if `tdd="true"` → TDD execution. Implement with deviation rules + auth gates. Verify done criteria. Commit (see task_commit). Track hash for Summary.
-   - `type="checkpoint:*"`: STOP → checkpoint_protocol → wait for user → continue only after confirmation.
-
-   **After each task completes successfully (WIRE-04 board transition):**
-   - Post task completion details on the task sub-issue:
-     ```bash
-     node ~/.claude/maxsim/bin/maxsim-tools.cjs github post-completion --issue-number $TASK_SUB_ISSUE_NUMBER --commit-sha "$TASK_COMMIT" --files-changed "file1.ts,file2.ts"
-     ```
-   - Close the task sub-issue and move to "Done":
-     ```bash
-     node ~/.claude/maxsim/bin/maxsim-tools.cjs github close-issue --issue-number $TASK_SUB_ISSUE_NUMBER
-     node ~/.claude/maxsim/bin/maxsim-tools.cjs github move-issue --issue-number $TASK_SUB_ISSUE_NUMBER --status "Done"
-     ```
-
-3. Run `<verification>` checks
-4. Confirm `<success_criteria>` met
-5. Document deviations in Summary
-</step>
-
-<authentication_gates>
-
-## Authentication Gates
-
-Auth errors during execution are NOT failures — they're expected interaction points.
-
-**Indicators:** "Not authenticated", "Unauthorized", 401/403, "Please run {tool} login", "Set {ENV_VAR}"
-
-**Protocol:**
-1. Recognize auth gate (not a bug)
-2. STOP task execution
-3. Create dynamic checkpoint:human-action with exact auth steps
-4. Wait for user to authenticate
-5. Verify credentials work
-6. Retry original task
-7. Continue normally
-
-**Example:** `vercel --yes` → "Not authenticated" → checkpoint asking user to `vercel login` → verify with `vercel whoami` → retry deploy → continue
-
-**In Summary:** Document as normal flow under "## Authentication Gates", not as deviations.
-
-</authentication_gates>
-
-<deviation_rules>
-
-## Deviation Rules
-
-You WILL discover unplanned work. Apply automatically, track all for Summary.
-
-| Rule | Trigger | Action | Permission |
-|------|---------|--------|------------|
-| **1: Bug** | Broken behavior, errors, wrong queries, type errors, security vulns, race conditions, leaks | Fix → test → verify → track `[Rule 1 - Bug]` | Auto |
-| **2: Missing Critical** | Missing essentials: error handling, validation, auth, CSRF/CORS, rate limiting, indexes, logging | Add → test → verify → track `[Rule 2 - Missing Critical]` | Auto |
-| **3: Blocking** | Prevents completion: missing deps, wrong types, broken imports, missing env/config/files, circular deps | Fix blocker → verify proceeds → track `[Rule 3 - Blocking]` | Auto |
-| **4: Architectural** | Structural change: new DB table, schema change, new service, switching libs, breaking API, new infra | STOP → present decision (below) → track `[Rule 4 - Architectural]` | Ask user |
-
-**Rule 4 format:**
+node ~/.claude/maxsim/bin/maxsim-tools.cjs github move-issue \
+  --issue-number {task_issue_number} --status "In Progress"
 ```
-⚠️ Architectural Decision Needed
-
-Current task: [task name]
-Discovery: [what prompted this]
-Proposed change: [modification]
-Why needed: [rationale]
-Impact: [what this affects]
-Alternatives: [other approaches]
 
-Proceed with proposed change? (yes / different approach / defer)
+Display:
 ```
-
-**Priority:** Rule 4 (STOP) > Rules 1-3 (auto) > unsure → Rule 4
-**Edge cases:** missing validation → R2 | null crash → R1 | new table → R4 | new column → R1/2
-**Heuristic:** Affects correctness/security/completion? → R1-3. Maybe? → R4.
-
-</deviation_rules>
-
-<deviation_documentation>
-
-## Documenting Deviations
-
-Summary MUST include deviations section. None? → `## Deviations from Plan\n\nNone - plan executed exactly as written.`
-
-Per deviation: **[Rule N - Category] Title** — Found during: Task X | Issue | Fix | Files modified | Verification | Commit hash
-
-End with: **Total deviations:** N auto-fixed (breakdown). **Impact:** assessment.
-
-</deviation_documentation>
-
-<tdd_plan_execution>
-## TDD Execution
-
-For `type: tdd` plans — RED-GREEN-REFACTOR:
-
-1. **Infrastructure** (first TDD plan only): detect project, install framework, config, verify empty suite
-2. **RED:** Read `<behavior>` → failing test(s) → run (MUST fail) → commit: `test({phase}-{plan}): add failing test for [feature]`
-3. **GREEN:** Read `<implementation>` → minimal code → run (MUST pass) → commit: `feat({phase}-{plan}): implement [feature]`
-4. **REFACTOR:** Clean up → tests MUST pass → commit: `refactor({phase}-{plan}): clean up [feature]`
-
-Errors: RED doesn't fail → investigate test/existing feature. GREEN doesn't pass → debug, iterate. REFACTOR breaks → undo.
-
-See `~/.claude/maxsim/references/tdd.md` for structure.
-</tdd_plan_execution>
-
-<task_commit>
-## Task Commit Protocol
-
-After each task (verification passed, done criteria met), commit immediately.
-
-**1. Check:** `git status --short`
-
-**2. Stage individually** (NEVER `git add .` or `git add -A`):
-```bash
-git add src/api/auth.ts
-git add src/types/user.ts
-```
-
-**3. Commit type:**
-
-| Type | When | Example |
-|------|------|---------|
-| `feat` | New functionality | feat(08-02): create user registration endpoint |
-| `fix` | Bug fix | fix(08-02): correct email validation regex |
-| `test` | Test-only (TDD RED) | test(08-02): add failing test for password hashing |
-| `refactor` | No behavior change (TDD REFACTOR) | refactor(08-02): extract validation to helper |
-| `perf` | Performance | perf(08-02): add database index |
-| `docs` | Documentation | docs(08-02): add API docs |
-| `style` | Formatting | style(08-02): format auth module |
-| `chore` | Config/deps | chore(08-02): add bcrypt dependency |
-
-**4. Format:** `{type}({phase}-{plan}): {description}` with bullet points for key changes.
-
-**5. Record hash:**
-```bash
-TASK_COMMIT=$(git rev-parse --short HEAD)
-TASK_COMMITS+=("Task ${TASK_NUM}: ${TASK_COMMIT}")
+## Task {N}/{total}: {task_name}
+Files: {files_to_modify}
 ```
 
-</task_commit>
-
-<step name="checkpoint_protocol">
-On `type="checkpoint:*"`: automate everything possible first. Checkpoints are for verification/decisions only.
-
-Display: `CHECKPOINT: [Type]` box → Progress {X}/{Y} → Task name → type-specific content → `YOUR ACTION: [signal]`
-
-| Type | Content | Resume signal |
-|------|---------|---------------|
-| human-verify (90%) | What was built + verification steps (commands/URLs) | "approved" or describe issues |
-| decision (9%) | Decision needed + context + options with pros/cons | "Select: option-id" |
-| human-action (1%) | What was automated + ONE manual step + verification plan | "done" |
+### 3b — Implement task
 
-After response: verify if specified. Pass → continue. Fail → inform, wait. WAIT for user — do NOT hallucinate completion.
+Read all files listed in the task's `files_to_modify` list using the Read tool.
 
-See ~/.claude/maxsim/references/checkpoints.md for details.
-</step>
+Implement the changes required by the task description and acceptance criteria.
 
-<step name="checkpoint_return_for_orchestrator">
-When spawned via Task and hitting checkpoint: return structured state (cannot interact with user directly).
+Conventions (from CLAUDE.md if present, otherwise defaults):
+- Follow existing code style and patterns
+- No dead code, no placeholder comments
+- All logic must be complete and functional (no TODOs)
+- Handle errors explicitly
+- Write tests if the task requires them or if the project has a test suite
 
-**Required return:** 1) Completed Tasks table (hashes + files) 2) Current Task (what's blocking) 3) Checkpoint Details (user-facing content) 4) Awaiting (what's needed from user)
+### 3c — Verify before commit
 
-Orchestrator parses → presents to user → spawns fresh continuation with your completed tasks state. You will NOT be resumed. In main context: use checkpoint_protocol above.
-</step>
-
-<step name="verification_failure_gate">
-If verification fails: STOP. Present: "Verification failed for Task [X]: [name]. Expected: [criteria]. Actual: [result]." Options: Retry | Skip (mark incomplete) | Stop (investigate). If skipped → SUMMARY "Issues Encountered".
-
-**On review failure (WIRE-07):** If the task sub-issue was already moved to "Done", reopen and move back:
+Run tests relevant to the changed files:
 ```bash
-node ~/.claude/maxsim/bin/maxsim-tools.cjs github reopen-issue --issue-number $TASK_SUB_ISSUE_NUMBER
-node ~/.claude/maxsim/bin/maxsim-tools.cjs github move-issue --issue-number $TASK_SUB_ISSUE_NUMBER --status "In Progress"
-TMPFILE=$(mktemp)
-cat > "$TMPFILE" << 'BODY_EOF'
-## Review Failure
-
-Verification failed for this task. Moving back to In Progress.
-
-**Reason:** {failure details}
-**Action needed:** {what needs to be fixed}
-BODY_EOF
-node ~/.claude/maxsim/bin/maxsim-tools.cjs github post-comment --issue-number $TASK_SUB_ISSUE_NUMBER --body-file "$TMPFILE"
-```
-</step>
-
-<step name="record_completion_time">
-```bash
-PLAN_END_TIME=$(date -u +"%Y-%m-%dT%H:%M:%SZ")
-PLAN_END_EPOCH=$(date +%s)
-
-DURATION_SEC=$(( PLAN_END_EPOCH - PLAN_START_EPOCH ))
-DURATION_MIN=$(( DURATION_SEC / 60 ))
-
-if [[ $DURATION_MIN -ge 60 ]]; then
-  HRS=$(( DURATION_MIN / 60 ))
-  MIN=$(( DURATION_MIN % 60 ))
-  DURATION="${HRS}h ${MIN}m"
-else
-  DURATION="${DURATION_MIN} min"
+# Detect and run test runner
+if [ -f package.json ]; then
+  # Check for test script
+  node -e "const p=require('./package.json'); process.exit(p.scripts?.test ? 0 : 1)" && npm test 2>&1 | tail -20
 fi
 ```
-</step>
-
-<step name="generate_user_setup">
-If plan frontmatter contains a `user_setup` field: create `{phase}-USER-SETUP.md` using template `~/.claude/maxsim/templates/user-setup.md`. Per service: env vars table, account setup checklist, dashboard config, local dev notes, verification commands. Status "Incomplete". Set `USER_SETUP_CREATED=true`. If empty/missing: skip.
-
-Also post a reminder comment on the phase issue:
-```bash
-TMPFILE=$(mktemp)
-cat > "$TMPFILE" << 'BODY_EOF'
-## User Setup Required
-
-This plan created a USER-SETUP.md with manual steps needed before proceeding.
-
-**File:** {phase}-USER-SETUP.md
-**Services:** {list of services requiring setup}
-BODY_EOF
-node ~/.claude/maxsim/bin/maxsim-tools.cjs github post-comment --issue-number $PHASE_ISSUE_NUMBER --body-file "$TMPFILE"
-```
-</step>
 
-<step name="create_summary">
+Check for anti-patterns in changed files:
+- No `TODO`, `FIXME`, `XXX`, `HACK` comments introduced
+- No placeholder content ("coming soon", "will be here")
+- No stub functions (functions that only `console.log` or `return null`)
+- No empty error handlers
 
-## Post Summary as GitHub Comment (WIRE-02)
+If tests fail: diagnose and fix. Retry up to 2 times before reporting failure.
 
-Do NOT write a local SUMMARY.md file to `.planning/phases/`. Instead, build the summary content in memory and post it as a GitHub comment.
+### 3d — Commit atomically
 
-Build summary content using the same structure as `~/.claude/maxsim/templates/summary.md`:
+Stage only the specific files changed by this task (NEVER `git add .` or `git add -A`):
 
-**Content to include:**
-- `<!-- maxsim:type=summary -->` HTML marker at the top (required for detection)
-- Frontmatter: phase, plan, subsystem, tags | requires/provides/affects | tech-stack.added/patterns | key-files.created/modified | key-decisions | requirements-completed (MUST copy `requirements` array from plan frontmatter verbatim) | duration ($DURATION), completed ($PLAN_END_TIME date)
-- Title: `# Phase [X] Plan [Y]: [Name] Summary`
-- One-liner SUBSTANTIVE: "JWT auth with refresh rotation using jose library" not "Authentication implemented"
-- Duration, start/end times, task count, file count
-- Next: more plans → "Ready for {next-plan}" | last → "Phase complete, ready for transition"
-- Deviations section (required, even if "None")
-- Review Cycle section (added after review cycle completes)
-- Self-Check result
-
-Post to the phase issue:
 ```bash
-TMPFILE=$(mktemp)
-echo "$SUMMARY_CONTENT" > "$TMPFILE"
-node ~/.claude/maxsim/bin/maxsim-tools.cjs github post-comment --issue-number $PHASE_ISSUE_NUMBER --body-file "$TMPFILE" --type summary
-```
-
-Record the comment URL/ID as `SUMMARY_COMMENT_ID` for future reference.
-
-GitHub Issues is the source of truth -- summaries are always posted as GitHub comments.
-</step>
-
-<step name="review_cycle">
-## Execute-Review-Simplify-Review Cycle
-
-After implementation is complete and summary is drafted (not yet posted), run the full review cycle before posting. All four stages must pass before the plan is considered done.
-
-### Retry Counter Initialization
-
-```
-SPEC_ATTEMPTS=0
-CODE_ATTEMPTS=0
-SIMPLIFY_ATTEMPTS=0
-FINAL_ATTEMPTS=0
-MAX_REVIEW_ATTEMPTS=3
-REVIEW_ESCALATIONS=0
-REVIEW_ESCALATION_DETAILS=""
-REVIEW_CYCLE_START=$(date +%s)
-```
-
----
-
-### Stage 1: Spec Review — Verify implementation matches plan spec.
-
+git add {specific_files_changed_by_this_task}
+git commit -m "{type}({phase_number}-{plan_id}): {task_description_as_commit_message}"
+TASK_COMMIT=$(git rev-parse HEAD)
 ```
-SPEC_STAGE_START=$(date +%s)
-SPEC_RESULT="PENDING"
-```
-
-**Retry loop (max MAX_REVIEW_ATTEMPTS):**
-
-1. Increment `SPEC_ATTEMPTS`
-2. Spawn verifier:
-
-```
-Task(
-  subagent_type="verifier",
-  prompt="
-    ## Task: Review for spec compliance
-
-    ## Suggested Skills: verification-gates, evidence-collection
-
-    <objective>
-    Review plan {phase}-{plan} for spec compliance.
-    </objective>
-
-    <plan_spec>
-    Plan content: {plan_content_in_memory or plan file path: {phase_dir}/{phase}-{plan}-PLAN.md}
-    Summary content: {summary_content_in_memory}
-    </plan_spec>
-
-    <task_specs>
-    {For each task in the plan: task number, name, done criteria, files modified}
-    </task_specs>
-
-    <files_modified>
-    {List all files created/modified during execution from summary key-files}
-    </files_modified>
-
-    <instructions>
-    1. Read the plan content and extract every task requirement
-    2. For each requirement, verify the implementation exists and matches the spec
-    3. Check that nothing was added beyond scope
-    4. Report: PASS (all requirements met) or FAIL (list unmet requirements)
-    </instructions>
-  "
-)
-```
-
-3. **If PASS:** Set `SPEC_RESULT="PASS"`, record `SPEC_STAGE_END=$(date +%s)`, proceed to Stage 2.
-
-4. **If FAIL and SPEC_ATTEMPTS < MAX_REVIEW_ATTEMPTS:** Fix the unmet requirements identified by the verifier, re-stage and commit fixes (`fix({phase}-{plan}): address spec review findings`), update summary content in memory, then loop back to step 1.
 
-5. **If FAIL and SPEC_ATTEMPTS >= MAX_REVIEW_ATTEMPTS:** Escalate to user:
+Commit type conventions:
+- `feat` — new feature or capability
+- `fix` — bug fix
+- `refactor` — code restructure without behavior change
+- `test` — adding or updating tests
+- `docs` — documentation only
+- `chore` — tooling, config, dependencies
 
-```markdown
-## Review Escalation: Spec Review Failed After 3 Attempts
+### 3e — Close task on board
 
-**Stage:** Spec Review
-**Attempts:** 3/3
-**Last failure reason:** {failure details from verifier}
-
-The spec review has failed 3 times. This may indicate a fundamental mismatch between the plan spec and the implementation.
-
-**Options:**
-1. Fix manually and type "retry" to re-run spec review (resets attempt counter)
-2. Type "override" to skip spec review (will be flagged in summary)
-3. Type "abort" to stop execution
-```
-
-Wait for user response:
-- **"retry":** Reset `SPEC_ATTEMPTS=0`, loop back to step 1.
-- **"override":** Set `SPEC_RESULT="OVERRIDDEN"`, increment `REVIEW_ESCALATIONS`, record `SPEC_STAGE_END=$(date +%s)`, proceed to Stage 2.
-- **"abort":** Stop execution, post partial summary with review status.
-
----
-
-### Stage 2: Code Review — Check code quality, security, error handling.
-
-```
-CODE_STAGE_START=$(date +%s)
-CODE_RESULT="PENDING"
+Move task sub-issue to Done and close it:
+```bash
+node ~/.claude/maxsim/bin/maxsim-tools.cjs github move-issue \
+  --issue-number {task_issue_number} --status "Done"
+node ~/.claude/maxsim/bin/maxsim-tools.cjs github close-issue \
+  --issue-number {task_issue_number}
 ```
 
-**Retry loop (max MAX_REVIEW_ATTEMPTS):**
-
-1. Increment `CODE_ATTEMPTS`
-2. Spawn verifier:
-
+Record task result:
 ```
-Task(
-  subagent_type="verifier",
-  prompt="
-    ## Task: Review code quality
-
-    ## Suggested Skills: code-review
-
-    <objective>
-    Review plan {phase}-{plan} code quality. Spec compliance already verified.
-    </objective>
-
-    <files_modified>
-    {List all files created/modified during execution from summary key-files}
-    </files_modified>
-
-    <instructions>
-    1. Read CLAUDE.md for project conventions
-    2. Review every modified file for: correctness, conventions, error handling, security, maintainability
-    3. Categorize findings: BLOCKER (must fix), HIGH (should fix), MEDIUM (file for follow-up)
-    4. Report: APPROVED (no blockers/high) or BLOCKED (list blocking issues)
-    </instructions>
-  "
-)
+Task {N}: {task_name} — DONE
+Commit: {TASK_COMMIT}
+Files: {files_changed}
 ```
 
-3. **If APPROVED:** Set `CODE_RESULT="APPROVED"`, record `CODE_STAGE_END=$(date +%s)`, proceed to Stage 3.
-
-4. **If BLOCKED and CODE_ATTEMPTS < MAX_REVIEW_ATTEMPTS:** Fix all blocker and high-severity issues, re-stage and commit fixes (`fix({phase}-{plan}): address code review findings`), then loop back to step 1.
+### 3f — Repeat for next task
 
-5. **If BLOCKED and CODE_ATTEMPTS >= MAX_REVIEW_ATTEMPTS:** Escalate to user:
+Pass only the commit hash and files-modified list to the next task context. Do NOT carry forward full diffs or task agent conversations.
 
-```markdown
-## Review Escalation: Code Review Failed After 3 Attempts
+## Step 4 — Execute Tasks (Pattern B — Segmented)
 
-**Stage:** Code Review
-**Attempts:** 3/3
-**Last failure reason:** {blocking issues from verifier}
+For plans with checkpoint markers:
 
-The code review has been blocked 3 times. Remaining issues may require architectural changes or user guidance.
+Execute each segment autonomously (tasks before the checkpoint). At each checkpoint:
+- Pause and surface the checkpoint to the orchestrator
+- Wait for orchestrator/user response before continuing
+- Resume with user feedback as additional context
 
-**Options:**
-1. Fix manually and type "retry" to re-run code review (resets attempt counter)
-2. Type "override" to skip code review (will be flagged in summary)
-3. Type "abort" to stop execution
-```
+After all segments complete, aggregate results for the summary step.
 
-Wait for user response:
-- **"retry":** Reset `CODE_ATTEMPTS=0`, loop back to step 1.
-- **"override":** Set `CODE_RESULT="OVERRIDDEN"`, increment `REVIEW_ESCALATIONS`, record `CODE_STAGE_END=$(date +%s)`, proceed to Stage 3.
-- **"abort":** Stop execution, post partial summary with review status.
+## Step 5 — Self-Check
 
-**On review failure -- reopen affected task sub-issues (WIRE-07):**
+Before posting the summary, run self-verification:
 
-If the code review identifies failures tied to specific tasks, and those task sub-issues were already closed:
 ```bash
-node ~/.claude/maxsim/bin/maxsim-tools.cjs github reopen-issue --issue-number $TASK_SUB_ISSUE_NUMBER
-node ~/.claude/maxsim/bin/maxsim-tools.cjs github move-issue --issue-number $TASK_SUB_ISSUE_NUMBER --status "In Progress"
-TMPFILE=$(mktemp)
-cat > "$TMPFILE" << 'BODY_EOF'
-## Code Review Failure
-
-This task's code review was blocked. Moving back to In Progress.
-
-**Blocking issues:**
-{list of blocker/high issues from verifier}
-BODY_EOF
-node ~/.claude/maxsim/bin/maxsim-tools.cjs github post-comment --issue-number $TASK_SUB_ISSUE_NUMBER --body-file "$TMPFILE"
-```
-
----
+# Verify key files from task list exist on disk
+for file in {key_files_created_or_modified}; do
+  [ -f "$file" ] && echo "EXISTS: $file" || echo "MISSING: $file"
+done
 
-### Stage 3: Simplify — Config-gated, spawn 3 parallel reviewers.
+# Verify commits exist for this plan
+git log --oneline --all --grep="{phase_number}-{plan_id}" 2>/dev/null
 
-**Config gate check:**
-```bash
-SIMPLIFY_ENABLED=$(node ~/.claude/maxsim/bin/maxsim-tools.cjs config-get review.simplify_review 2>/dev/null || echo "true")
-```
-
-**If `SIMPLIFY_ENABLED` is "false":** Skip Stage 3 and Stage 4 entirely. Log: "Simplify stage skipped (disabled in config)". Set `SIMPLIFY_RESULT="SKIPPED"`, `FINAL_RESULT="N/A"`. Proceed to review cycle tracking.
-
-**If `SIMPLIFY_ENABLED` is "true" (default):**
-
-```
-SIMPLIFY_STAGE_START=$(date +%s)
-SIMPLIFY_RESULT="PENDING"
-SIMPLIFY_ATTEMPTS=0
-```
-
-**Reviewer 1 -- Code Reuse:**
-
-```
-Task(
-  subagent_type="verifier",
-  prompt="
-    ## Task: Review for code reuse opportunities
-
-    ## Suggested Skills: code-review
-
-    <objective>
-    Review plan {phase}-{plan} for code reuse opportunities.
-    </objective>
-
-    <files_to_review>
-    {List all files created/modified during execution from summary key-files}
-    </files_to_review>
-
-    <instructions>
-    1. Scan all changed files for duplicated patterns
-    2. Cross-reference against existing shared utilities and helpers
-    3. Flag any logic that appears 3+ times without extraction
-    4. Check if existing utilities could replace new code
-    5. Output: list of reuse opportunities with file paths and line ranges
-    6. Verdict: CLEAN (no issues) or ISSUES_FOUND (list)
-    </instructions>
-  "
-)
-```
-
-**Reviewer 2 -- Code Quality:**
-
-```
-Task(
-  subagent_type="verifier",
-  prompt="
-    ## Task: Review for code quality issues
-
-    ## Suggested Skills: code-review
-
-    <objective>
-    Review plan {phase}-{plan} for code quality issues.
-    </objective>
-
-    <files_to_review>
-    {List all files created/modified during execution from summary key-files}
-    </files_to_review>
-
-    <instructions>
-    1. Check naming consistency with codebase conventions
-    2. Verify error handling covers all external calls
-    3. Look for dead code: unused imports, unreachable branches, commented-out code
-    4. Check for unnecessary abstractions or premature generalizations
-    5. Output: list of quality issues categorized by severity (BLOCKER/HIGH/MEDIUM)
-    6. Verdict: CLEAN (no issues) or ISSUES_FOUND (list)
-    </instructions>
-  "
-)
+# Verify all task sub-issues are closed
+node ~/.claude/maxsim/bin/maxsim-tools.cjs github list-sub-issues \
+  --phase-issue-number {phase_issue_number}
 ```
 
-**Reviewer 3 -- Efficiency:**
+Self-check result:
+- All key files exist + all commits found + all sub-issues closed → `## Self-Check: PASSED`
+- Any miss → `## Self-Check: FAILED — {reason}`
 
-```
-Task(
-  subagent_type="verifier",
-  prompt="
-    ## Task: Review for efficiency issues
-
-    ## Suggested Skills: code-review
-
-    <objective>
-    Review plan {phase}-{plan} for efficiency issues.
-    </objective>
-
-    <files_to_review>
-    {List all files created/modified during execution from summary key-files}
-    </files_to_review>
-
-    <instructions>
-    1. Find over-engineered solutions (parametrization serving one case, generic interfaces with one implementor)
-    2. Identify repeated computations that could be cached or hoisted
-    3. Check for unnecessary allocations in hot paths
-    4. Look for redundant data transformations
-    5. Output: list of efficiency issues with suggested removals
-    6. Verdict: CLEAN (no issues) or ISSUES_FOUND (list)
-    </instructions>
-  "
-)
-```
+## Step 6 — Post Summary to GitHub
 
-**Consolidation (after all 3 reviewers complete):**
+Build summary content and post as a comment on the phase issue.
 
-After all three reviewers report:
-- If ALL returned CLEAN: Set `SIMPLIFY_RESULT="CLEAN"`, record `SIMPLIFY_STAGE_END=$(date +%s)`, skip Stage 4. Set `FINAL_RESULT="N/A"`.
-- If ANY returned ISSUES_FOUND:
-  1. Increment `SIMPLIFY_ATTEMPTS`
-  2. Merge findings into deduplicated list
-  3. Spawn single executor to apply fixes:
+The summary comment IS the completion record. No local SUMMARY.md file is written.
 
-```
-Task(
-  subagent_type="executor",
-  prompt="
-    <objective>
-    Apply simplification fixes for plan {phase}-{plan} based on reviewer findings.
-    </objective>
-
-    <findings>
-    {Merged deduplicated list of ISSUES_FOUND from all reviewers — BLOCKER and HIGH first, skip speculative optimizations}
-    </findings>
-
-    <files_to_review>
-    {List all files created/modified during execution from summary key-files}
-    </files_to_review>
-
-    <instructions>
-    1. Apply fixes for all actionable items (BLOCKER and HIGH severity first)
-    2. Skip speculative optimizations that lack clear evidence
-    3. Run tests to confirm fixes do not break anything
-    4. Report: FIXED (issues found and resolved) or BLOCKED (cannot fix without architectural change)
-    </instructions>
-  "
-)
-```
-
-4. **If FIXED:** Set `SIMPLIFY_RESULT="FIXED"`, record `SIMPLIFY_STAGE_END=$(date +%s)`, proceed to Stage 4.
-
-5. **If BLOCKED and SIMPLIFY_ATTEMPTS < MAX_REVIEW_ATTEMPTS:** Re-run the 3 reviewers to find alternative fixes, then loop.
-
-6. **If BLOCKED and SIMPLIFY_ATTEMPTS >= MAX_REVIEW_ATTEMPTS:** Escalate to user:
-
-```markdown
-## Review Escalation: Simplify Stage Blocked After 3 Attempts
-
-**Stage:** Simplify
-**Attempts:** 3/3
-**Last failure reason:** {architectural issues from executor}
-
-The simplify stage has been blocked 3 times. Remaining issues require architectural changes.
-
-**Options:**
-1. Fix manually and type "retry" to re-run simplify (resets attempt counter)
-2. Type "override" to skip simplify (will be flagged in summary)
-3. Type "abort" to stop execution
-```
-
-Wait for user response:
-- **"retry":** Reset `SIMPLIFY_ATTEMPTS=0`, re-run 3 reviewers.
-- **"override":** Set `SIMPLIFY_RESULT="OVERRIDDEN"`, increment `REVIEW_ESCALATIONS`, record `SIMPLIFY_STAGE_END=$(date +%s)`, skip Stage 4. Set `FINAL_RESULT="N/A"`.
-- **"abort":** Stop execution, post partial summary with review status.
+```bash
+PLAN_END_EPOCH=$(date +%s)
+PLAN_DURATION=$(( (PLAN_END_EPOCH - PLAN_START_EPOCH) / 60 ))m
 
+SUMMARY_FILE=$(mktemp)
+cat > "$SUMMARY_FILE" << 'SUMMARY_EOF'
 ---
-
-### Stage 4: Final Review — If simplify made changes, one more code review pass.
-
-Run this stage ONLY if Stage 3 reported FIXED (i.e., simplify found and applied fixes). If Stage 3 was CLEAN, SKIPPED, or OVERRIDDEN: set `FINAL_RESULT="N/A"`, skip to review cycle tracking.
-
-```
-FINAL_STAGE_START=$(date +%s)
-FINAL_RESULT="PENDING"
-```
-
-**Retry loop (max MAX_REVIEW_ATTEMPTS):**
-
-1. Increment `FINAL_ATTEMPTS`
-2. Spawn verifier:
-
-```
-Task(
-  subagent_type="verifier",
-  prompt="
-    ## Task: Final review after simplification
-
-    ## Suggested Skills: code-review
-
-    <objective>
-    Final review pass after simplification changes on plan {phase}-{plan}.
-    Verify simplification fixes did not introduce regressions.
-    </objective>
-
-    <context>
-    Previous code review: APPROVED
-    Simplification: FIXED — changes were applied
-    This is a focused re-review of simplification changes only.
-    </context>
-
-    <files_modified>
-    {List files changed by simplification stage}
-    </files_modified>
-
-    <instructions>
-    1. Review only the changes made during simplification
-    2. Verify no regressions were introduced
-    3. Check that simplification changes follow project conventions
-    4. Report: APPROVED or BLOCKED (list issues)
-    </instructions>
-  "
-)
-```
-
-3. **If APPROVED:** Set `FINAL_RESULT="APPROVED"`, record `FINAL_STAGE_END=$(date +%s)`, proceed to review cycle tracking.
-
-4. **If BLOCKED and FINAL_ATTEMPTS < MAX_REVIEW_ATTEMPTS:** Fix issues, re-stage and commit fixes (`fix({phase}-{plan}): address final review findings`), then loop back to step 1.
-
-5. **If BLOCKED and FINAL_ATTEMPTS >= MAX_REVIEW_ATTEMPTS:** Escalate to user:
-
-```markdown
-## Review Escalation: Final Review Blocked After 3 Attempts
-
-**Stage:** Final Review
-**Attempts:** 3/3
-**Last failure reason:** {blocking issues from verifier}
-
-The final review has been blocked 3 times after simplification changes.
-
-**Options:**
-1. Fix manually and type "retry" to re-run final review (resets attempt counter)
-2. Type "override" to skip final review (will be flagged in summary)
-3. Type "abort" to stop execution
-```
-
-Wait for user response:
-- **"retry":** Reset `FINAL_ATTEMPTS=0`, loop back to step 1.
-- **"override":** Set `FINAL_RESULT="OVERRIDDEN"`, increment `REVIEW_ESCALATIONS`, record `FINAL_STAGE_END=$(date +%s)`, proceed to review cycle tracking.
-- **"abort":** Stop execution, post partial summary with review status.
-
+phase: {phase_number}
+plan: {plan_id}
+objective: {objective}
+wave: {wave_number}
+completed: {ISO timestamp}
+duration: {PLAN_DURATION}
+key_files_created:
+{list of files created}
+key_files_modified:
+{list of files modified}
+key_decisions:
+{list of any architectural decisions made during execution}
 ---
 
-### Review Cycle Tracking
-
-```
-REVIEW_CYCLE_END=$(date +%s)
-REVIEW_CYCLE_TOTAL=$(( REVIEW_CYCLE_END - REVIEW_CYCLE_START ))
-```
-
-Calculate per-stage durations:
-```
-SPEC_DURATION=$(( SPEC_STAGE_END - SPEC_STAGE_START ))
-CODE_DURATION=$(( CODE_STAGE_END - CODE_STAGE_START ))
-# Only if simplify ran:
-SIMPLIFY_DURATION=$(( SIMPLIFY_STAGE_END - SIMPLIFY_STAGE_START ))  # or "N/A" if skipped
-# Only if final review ran:
-FINAL_DURATION=$(( FINAL_STAGE_END - FINAL_STAGE_START ))  # or "N/A" if skipped
-```
-
-Add review cycle results to the summary content in memory (under a `## Review Cycle` section):
-```markdown
-## Review Cycle
+## {Plan ID}: {Objective}
 
-| Stage | Result | Attempts | Duration | Findings |
-|-------|--------|----------|----------|----------|
-| Spec Review | {PASS|FAIL|OVERRIDDEN} | {SPEC_ATTEMPTS}/3 | {SPEC_DURATION}s | {summary or "All requirements met"} |
-| Code Review | {APPROVED|BLOCKED|OVERRIDDEN} | {CODE_ATTEMPTS}/3 | {CODE_DURATION}s | {summary or "No blocking issues"} |
-| Simplify | {CLEAN|FIXED|BLOCKED|SKIPPED|OVERRIDDEN} | {SIMPLIFY_ATTEMPTS}/3 | {SIMPLIFY_DURATION}s | {summary or "N/A"} |
-| Final Review | {APPROVED|BLOCKED|SKIPPED|N/A|OVERRIDDEN} | {FINAL_ATTEMPTS}/3 | {FINAL_DURATION}s | {summary or "N/A"} |
+{One-liner: substantive description of what was built and how it works}
 
-**Total review time:** {REVIEW_CYCLE_TOTAL}s
-**Escalations:** {REVIEW_ESCALATIONS} ({REVIEW_ESCALATION_DETAILS or "None"})
-```
+## Task Execution
 
-Now post the complete summary (with review cycle included) to GitHub:
-```bash
-TMPFILE=$(mktemp)
-echo "$COMPLETE_SUMMARY_CONTENT" > "$TMPFILE"
-node ~/.claude/maxsim/bin/maxsim-tools.cjs github post-comment --issue-number $PHASE_ISSUE_NUMBER --body-file "$TMPFILE" --type summary
-```
+| Task | Name | Status | Commit | Notes |
+|------|------|--------|--------|-------|
+| 1 | {name} | PASS | {hash} | |
+| 2 | {name} | PASS | {hash} | {deviation if any} |
 
-After posting, commit any review-cycle metadata:
-```bash
-node ~/.claude/maxsim/bin/maxsim-tools.cjs commit "docs({phase}-{plan}): add review cycle results" --files .planning/STATE.md .planning/ROADMAP.md
-```
-</step>
+## Deviations
+{List any departures from the plan spec, with rationale. "None." if no deviations.}
 
-<step name="evidence_gate">
-**Gate G6: Evidence Completeness** — Summary must have evidence for each requirement.
+## Issues Encountered
+{List any problems hit during execution and how they were resolved. "None." if clean.}
 
-After the review cycle and before finalizing, validate that the summary content contains requirement evidence for all requirements in the plan's frontmatter.
+{self_check_result}
+<!-- maxsim:type=summary -->
+SUMMARY_EOF
 
-```bash
-if [ -n "$REQ_IDS" ]; then
-  # Write summary content to a temp file for verification
-  echo "$SUMMARY_CONTENT" > /tmp/summary-check.md
-  G6_RESULT=$(node ~/.claude/maxsim/bin/maxsim-tools.cjs verify evidence-completeness "/tmp/summary-check.md" $REQ_IDS)
-  G6_VALID=$(echo "$G6_RESULT" | node -e "process.stdin.on('data',d=>console.log(JSON.parse(d).valid))")
-  if [ "$G6_VALID" != "true" ]; then
-    G6_MISSING=$(echo "$G6_RESULT" | node -e "process.stdin.on('data',d=>console.log(JSON.parse(d).missing_evidence.join(', ')))")
-    echo "GATE G6: Missing requirement evidence for: $G6_MISSING"
-    echo "Add Requirement Evidence rows to summary for these requirements before posting."
-    # Present as warning with instruction to fix — executor should add evidence
-  fi
-fi
+node ~/.claude/maxsim/bin/maxsim-tools.cjs github post-comment \
+  --issue-number {phase_issue_number} \
+  --body-file "$SUMMARY_FILE" \
+  --type summary
 ```
-</step>
 
-<step name="update_current_position">
-Update STATE.md using maxsim-tools:
+## Step 7 — Update State Files
 
 ```bash
-# Advance plan counter (handles last-plan edge case)
+# Advance plan counter
 node ~/.claude/maxsim/bin/maxsim-tools.cjs state advance-plan
 
-# Recalculate progress bar from disk state
+# Recalculate progress
 node ~/.claude/maxsim/bin/maxsim-tools.cjs state update-progress
 
-# Record execution metrics
+# Record session metrics
 node ~/.claude/maxsim/bin/maxsim-tools.cjs state record-metric \
-  --phase "${PHASE}" --plan "${PLAN}" --duration "${DURATION}" \
-  --tasks "${TASK_COUNT}" --files "${FILE_COUNT}"
-```
-</step>
+  --phase "{PHASE_NUMBER}" --plan "{PLAN_ID}" --duration "{PLAN_DURATION}" \
+  --tasks "{TASK_COUNT}" --files "{FILE_COUNT}"
 
-<step name="extract_decisions_and_issues">
-From summary content: Extract decisions and add to STATE.md:
-
-```bash
-# Add each decision from summary key-decisions
-# Prefer file inputs for shell-safe text (preserves `$`, `*`, etc. exactly)
-node ~/.claude/maxsim/bin/maxsim-tools.cjs state add-decision \
-  --phase "${PHASE}" --summary-file "${DECISION_TEXT_FILE}" --rationale-file "${RATIONALE_FILE}"
-
-# Add blockers if any found
-node ~/.claude/maxsim/bin/maxsim-tools.cjs state add-blocker --text-file "${BLOCKER_TEXT_FILE}"
-```
-</step>
-
-<step name="update_session_continuity">
-Update session info using maxsim-tools:
-
-```bash
+# Record session stop position
 node ~/.claude/maxsim/bin/maxsim-tools.cjs state record-session \
-  --stopped-at "Completed ${PHASE}-${PLAN}-PLAN.md" \
+  --stopped-at "Completed {PHASE_NUMBER}-{PLAN_ID} (execute-plan)" \
   --resume-file "None"
-```
-
-Keep STATE.md under 150 lines.
-</step>
-
-<step name="issues_review_gate">
-If summary "Issues Encountered" ≠ "None": yolo → log and continue. Interactive → present issues, wait for acknowledgment.
-</step>
-
-<step name="update_roadmap">
-```bash
-node ~/.claude/maxsim/bin/maxsim-tools.cjs roadmap update-plan-progress "${PHASE}"
-```
-Counts completed task sub-issues (or PLAN vs SUMMARY files in fallback mode) on disk. Updates progress table row with correct count and status (`In Progress` or `Complete` with date).
-</step>
-
-<step name="update_requirements">
-Mark completed requirements from the plan's frontmatter `requirements:` field:
-
-```bash
-node ~/.claude/maxsim/bin/maxsim-tools.cjs requirements mark-complete ${REQ_IDS}
-```
-
-Extract requirement IDs from the plan frontmatter (e.g., `requirements: [AUTH-01, AUTH-02]`). If no requirements field, skip.
-</step>
-
-<step name="git_commit_metadata">
-Task code already committed per-task. Commit plan metadata:
-
-```bash
-node ~/.claude/maxsim/bin/maxsim-tools.cjs commit "docs({phase}-{plan}): complete [plan-name] plan" --files .planning/STATE.md .planning/ROADMAP.md .planning/REQUIREMENTS.md
-```
-
-Note: No local SUMMARY.md is committed -- summary was posted to GitHub as a comment.
-</step>
-
-<step name="post_verification_to_github">
-
-## Post Verification Results as GitHub Comment
-
-If the plan includes verification criteria (from `<verification>` section of plan content):
-
-Build verification results content in memory with `<!-- maxsim:type=verification -->` marker.
-
-```bash
-TMPFILE=$(mktemp)
-cat > "$TMPFILE" << 'BODY_EOF'
-<!-- maxsim:type=verification -->
-## Plan {plan_number} Verification
-
-{verification_results}
-
-**Status:** {passed|failed}
-**Timestamp:** {ISO timestamp}
-BODY_EOF
-node ~/.claude/maxsim/bin/maxsim-tools.cjs github post-comment --issue-number $PHASE_ISSUE_NUMBER --body-file "$TMPFILE" --type verification
-```
 
-If the plan includes UAT criteria (from `<uat>` section):
-
-```bash
-TMPFILE=$(mktemp)
-cat > "$TMPFILE" << 'BODY_EOF'
-<!-- maxsim:type=uat -->
-## Plan {plan_number} UAT
-
-{uat_criteria_and_results}
-
-**Status:** {pending|passed|failed}
-**Timestamp:** {ISO timestamp}
-BODY_EOF
-node ~/.claude/maxsim/bin/maxsim-tools.cjs github post-comment --issue-number $PHASE_ISSUE_NUMBER --body-file "$TMPFILE" --type uat
-```
-
-GitHub Issues is the source of truth -- verification and UAT results are always posted as GitHub comments.
-</step>
-
-<step name="update_codebase_map">
-If .planning/codebase/ doesn't exist: skip.
-
-```bash
-FIRST_TASK=$(git log --oneline --grep="feat({phase}-{plan}):" --grep="fix({phase}-{plan}):" --grep="test({phase}-{plan}):" --reverse | head -1 | cut -d' ' -f1)
-git diff --name-only ${FIRST_TASK}^..HEAD 2>/dev/null
-```
-
-Update only structural changes: new src/ dir → STRUCTURE.md | deps → STACK.md | file pattern → CONVENTIONS.md | API client → INTEGRATIONS.md | config → STACK.md | renamed → update paths. Skip code-only/bugfix/content changes.
-
-```bash
-node ~/.claude/maxsim/bin/maxsim-tools.cjs commit "" --files .planning/codebase/*.md --amend
+# Update roadmap progress
+node ~/.claude/maxsim/bin/maxsim-tools.cjs roadmap update-plan-progress "{PHASE_NUMBER}"
 ```
-</step>
-
-<step name="offer_next">
-If `USER_SETUP_CREATED=true`: display `⚠️ USER SETUP REQUIRED` with path + env/config tasks at TOP.
 
-Check completion by querying the phase issue's task sub-issues:
+Commit planning metadata (task code was already committed per-task):
 ```bash
-node ~/.claude/maxsim/bin/maxsim-tools.cjs github list-sub-issues --phase-issue-number $PHASE_ISSUE_NUMBER
+node ~/.claude/maxsim/bin/maxsim-tools.cjs commit \
+  "docs({phase_number}-{plan_id}): complete plan execution" \
+  --files .planning/STATE.md .planning/ROADMAP.md .planning/REQUIREMENTS.md
 ```
 
-Count open vs closed sub-issues. Map closed count to plans complete.
+## Step 8 — Return Result
 
-| Condition | Route | Action |
-|-----------|-------|--------|
-| open task sub-issues remain | **A: More plans** | Find next incomplete plan (by open sub-issues or missing summary comment). Yolo: auto-continue. Interactive: show next plan, suggest `/maxsim:execute {phase}`. STOP here. |
-| all sub-issues closed, current < highest phase | **B: Phase done** | Show completion, suggest `/maxsim:plan {Z+1}`. |
-| all sub-issues closed, current = highest phase | **C: Milestone done** | Show banner, suggest `/maxsim:progress`. |
+The final line of output MUST be exactly one of:
+- `RESULT: PASS`
+- `RESULT: FAIL — {specific reason}`
 
-All routes: `/clear` first for fresh context.
-</step>
+Do not add text after the RESULT line.
 
 </process>
 
-<success_criteria>
+<verification_checklist>
+Before returning RESULT: PASS, confirm ALL of the following:
+- [ ] All tasks in the plan were executed
+- [ ] Each task has exactly one atomic commit with a conventional commit message
+- [ ] No task was committed with `git add .` or `git add -A`
+- [ ] Tests were run and pass (or no test runner detected)
+- [ ] No TODOs, stubs, or placeholder content introduced
+- [ ] All task sub-issues are closed on GitHub
+- [ ] Summary comment posted to phase issue with <!-- maxsim:type=summary -->
+- [ ] Self-check ran and result appended to summary
+- [ ] STATE.md and ROADMAP.md updated
+</verification_checklist>
+
+<failure_handling>
+- **Task tests fail after 2 retries:** Diagnose root cause. If cannot fix, return `RESULT: FAIL — task {N} tests failing: {error}`
+- **File cannot be found:** Check path relative to worktree root. If genuinely missing, create it as specified in the task.
+- **Sub-issue board transition fails:** Log the error, continue execution. Include note in summary.
+- **GitHub post-comment fails:** Retry once with exponential backoff. If still failing, write summary to `.planning/phases/{phase_dir}/{plan_id}-SUMMARY.md` as fallback and report error.
+- **Commit fails (merge conflict):** You are in an isolated worktree. Conflicts indicate unexpected overlap. Report immediately: `RESULT: FAIL — merge conflict in worktree: {files}`
+</failure_handling>
 
-- All tasks from plan completed
-- All verifications pass
-- USER-SETUP.md generated if user_setup in frontmatter
-- Summary posted as GitHub comment with `<!-- maxsim:type=summary -->` marker (not written to local SUMMARY.md)
-- Verification and UAT posted as GitHub comments with appropriate type markers
-- Task sub-issues moved: In Progress when started, Done when completed (WIRE-04)
-- On review failure: task sub-issues reopened and moved back to In Progress (WIRE-07)
-- STATE.md updated (position, decisions, issues, session)
-- ROADMAP.md updated
-- If codebase map exists: map updated with execution changes (or skipped if no significant changes)
-- If USER-SETUP.md created: prominently surfaced in completion output
+<success_criteria>
+- [ ] Plan content parsed from GitHub context (not from local PLAN.md)
+- [ ] External edit detection checked before execution begins
+- [ ] All tasks executed in order
+- [ ] Each task committed individually with conventional commit message
+- [ ] Board transitions: In Progress when task starts, Done+closed when task completes
+- [ ] Self-check ran and result recorded in summary
+- [ ] Summary posted as GitHub comment (<!-- maxsim:type=summary -->) — no local SUMMARY.md written
+- [ ] State files updated
+- [ ] Final output line is exactly RESULT: PASS or RESULT: FAIL — {reason}
 </success_criteria>