claude-devkit-cli 1.2.4 → 1.3.0

package/templates/.claude/commands/mf-plan.md

@@ -1,141 +1,501 @@
-Generate spec + test plan from description or existing spec.
+Generate spec with acceptance scenarios from description or existing spec.
 
 ## Determine mode
 
 Examine `$ARGUMENTS`:
 
-- **Mode A — Spec exists:** Argument is a file path → read spec, generate test plan.
-- **Mode B — No spec:** Argument is a description → create spec + test plan.
-- **Mode C — Update:** Argument mentions "update" or existing path → read existing, update surgically.
+- **Mode A — New spec:** Argument is a feature description AND directory
+  `docs/specs/<feature>/` does not exist → create new spec.
+- **Mode B — Add scenarios:** Argument is a path to an existing spec AND spec does not
+  contain `## Stories` section with AS-NNN IDs → read spec, add acceptance scenarios.
+- **Mode C — Update:** Argument is a path to an existing spec AND spec already contains
+  `## Stories` section with AS-NNN IDs → update flow (see Mode C section below).
+
+---
+
+## Directory Structure
+
+```
+docs/specs/
+  <feature>/
+    <feature>.md                    # current state — always read this file
+    snapshots/                      # version history
+      <YYYY-MM-DD>.md
+      <YYYY-MM-DD>-<REF>.md
+```
+
+- `<feature>.md` is the single source of truth. All spec reads start from this file.
+- `snapshots/` contains full copies at points in time. Immutable — never edit a snapshot.
+- When a feature is split, sub-specs live in the same directory:
+  ```
+  docs/specs/billing/
+    billing.md                # root spec or overview
+    billing-checkout.md       # sub-spec
+    billing-refund.md         # sub-spec
+    snapshots/
+  ```
 
 ---
 
 ## Phase 0: Codebase Awareness
 
-Before writing anything:
-1. Scan existing code in the feature area — what files, functions, types already exist?
-2. Check `docs/specs/` — is there already a spec for this or a related feature?
-3. Check `docs/test-plans/` — any overlap with existing plans?
-4. Identify project patterns — test framework, naming conventions, directory structure.
+Before writing anything, run this checklist:
+
+| # | Action | How |
+|---|--------|-----|
+| P0-1 | **Keyword scan** | Grep the codebase for 3-5 keywords from the feature description. Note matching files, functions, types. |
+| P0-2 | **Related specs** | List `docs/specs/` directories. Read the main spec of any related feature. Is there overlap? |
+| P0-3 | **Dependency scan** | In the feature area, check imports/dependencies. What modules does this code touch? |
+| P0-4 | **Reusable utilities** | Look for existing helpers, validators, formatters, shared types that the new feature could reuse. List candidates. |
+| P0-5 | **Project patterns** | Identify test framework, naming conventions, directory structure from existing code. |
+| P0-6 | **Change Log** | If the feature exists, read its Change Log to understand evolution. |
+| P0-7 | **Knowledge graph** | If `codebase-memory-mcp` is available, use `search_code`, `get_architecture`, and `trace_call_path` to discover related code, understand architecture context, and trace dependencies — faster and more thorough than manual grep. |
+
+If `codebase-memory-mcp` MCP server is connected, prefer it for P0-1, P0-3, P0-4 — it provides indexed search, architecture overview, and call path tracing that are more reliable than ad-hoc grep.
+
+Record findings as bullet points — carry them into Phase 2 (Data Model, Constraints) and Phase 3 (ambiguity check).
 
 Don't plan in a vacuum. A spec that ignores existing code creates conflicts.
 
 ---
 
-## Phase 1: Draft the Spec (Mode B only)
+## Phase 1: Scope & Split Assessment
+
+Before writing the spec, assess size.
+
+**Input:** Feature description from user (Mode A) or current spec (Mode B).
+Mode C does not run Phase 1 — it uses its own flow (see Mode C section).
+
+**Split rules:**
+
+| # | Condition | Action |
+|---|-----------|--------|
+| T1 | Feature has >7 expected stories | MUST split |
+| T2 | Feature has >20 expected AS | MUST split |
+| T3 | Stories belong to different domains (e.g. payment + notification) | SHOULD split |
+| T4 | A story can ship independently without depending on other stories | SHOULD split |
+| T5 | Stories share a data model or state machine | DO NOT split |
+| T6 | Splitting would duplicate >50% of context (entities, constraints) | DO NOT split |
+
+"MUST" = mandatory split, inform user.
+"SHOULD" = suggest split, ask user.
+"DO NOT" = keep together, unless user requests split.
+
+If splitting:
+- Create the feature directory, place sub-specs in the same directory.
+- Each sub-spec must be self-contained (own overview, relevant data model, constraints).
+- No sub-spec should depend on another sub-spec to be understood.
+
+---
+
+## Phase 2: Draft the Spec (Mode A + B)
+
+**Mode A:** Create a new spec at `docs/specs/<feature>/<feature>.md` using the template below. Include stories + acceptance scenarios.
+
+**Mode B:** Read existing spec, add `## Stories` section with AS following depth rules.
+
+### Spec Template
+
+```markdown
+# Spec: <Feature Name>
+
+**Created:** <$(date +%Y-%m-%d)>
+**Last updated:** <$(date +%Y-%m-%d)>
+**Status:** Draft | Active | Deprecated
+**Snapshot limit:** <N, optional — default 5>
+
+## Overview
+[what, why, who — 2-3 sentences]
+
+## Data Model
+[entities, attributes, relationships — if applicable]
+
+## Stories
+
+### S-001: <Story name> (P0)
+
+**Description:** [user story]
+**Source:** [optional: ticket/issue ref]
+
+**Acceptance Scenarios:**
+
+AS-001: <short description>
+- **Given:** [state]
+- **When:** [action]
+- **Then:** [expected]
+- **Data:** [test data]
+
+AS-002: <short description>
+- **Given:** [error state]
+- **When:** [action]
+- **Then:** [error handling]
+- **Data:** [edge case data]
+
+### S-002: <Story name> (P1)
+
+**Description:** [user story]
+**Source:** [optional]
+
+**Acceptance Scenarios:**
+
+AS-003: <short description>
+- **Given:** [state]
+- **When:** [action]
+- **Then:** [expected]
+
+### S-003: <Story name> (P2)
+
+**Description:** [user story]
+
+**Acceptance Scenarios:**
+
+AS-004: <short description>
+- [flow description + expected behavior]
+
+## Constraints & Invariants
+[rules that must ALWAYS hold]
+
+## Change Log
+
+| Date | Change | Ref |
+|------|--------|-----|
+| <$(date +%Y-%m-%d)> | Initial creation | -- |
+```
+
+### Acceptance Scenario Depth
+
+| Story priority | AS must contain | AS optional |
+|---------------|----------------|-------------|
+| P0 | Given + When + Then + Data + Setup | -- |
+| P1 | Given + When + Then | Data, Setup |
+| P2 | 1-2 line flow description + expected | Separate Given/When/Then |
+
+**AS rules:**
+- Every P0 story must have at least 1 happy path AS + 1 error path AS.
+- Every P1 story must have at least 1 happy path AS.
+- Every P2 story must have at least 1 AS.
+- No orphan AS — every AS belongs to exactly 1 story.
+
+Match depth to complexity. Simple CRUD = 3 stories. Complex auth = full template.
+
+### Writing Instructions
+
+When generating stories and acceptance scenarios:
+
+**DO:**
+- Write AS that test one specific behavior each. If it fails, the developer knows exactly what broke.
+- Use concrete values in Given/When/Then — `Given: user with balance $50` not `Given: user with some balance`.
+- Name edge cases explicitly — `AS-005: Payment with insufficient funds` not `AS-005: Payment error`.
+- Each AS should be independent — no AS depends on another running first.
+- Include the boundary — `Given: cart with 0 items` and `Given: cart with 999 items`, not just `Given: cart with items`.
+
+**DO NOT produce:**
+- Vague AS: "Test that the feature works" — every AS must specify Given, When, Then (or a concrete flow for P2).
+- Excessive AS: 30+ scenarios for simple CRUD — over-testing wastes time and creates maintenance burden.
+- Implementation-testing AS: "Test that the database query uses an index" — test behavior, not internals.
+- Duplicate AS: two scenarios verifying the same behavior with trivially different inputs.
+- Framework-testing AS: "Test that the router handles the path" — test YOUR logic, not the framework.
+
+### Spec Section Guidelines
+
+Include only sections that apply. Skip sections that add no value:
+- **Data Model** — skip if feature has no persistent data or entities.
+- **Constraints & Invariants** — skip if no rules must always hold.
+- Don't generate filler for sections that don't apply.
 
-Create at `docs/specs/<feature-name>.md`. Include these sections (skip any that don't apply):
+### Consistency Check (after drafting)
 
-- **Overview** — what, why, who. 2-3 sentences.
-- **Data Model** — entities, attributes, relationships (table format)
-- **Use Cases** — UC-NNN with actor, preconditions, flow, postconditions, error cases. Each use case contains:
-  - **FR-NNN** (Functional Requirements) — specific behaviors the system must exhibit
-  - **SC-NNN** (Success Criteria) — measurable non-functional targets (performance, limits)
-- **State Machine** — states and valid transitions (if applicable)
-- **Settings/Configuration** — configurable behavior and defaults
-- **Constraints & Invariants** — rules that must ALWAYS hold
-- **Error Handling** — how errors surface to users and are logged
-- **Security Considerations** — auth, authorization, data sensitivity
+Before showing the draft, verify:
 
-Match depth to complexity. Simple CRUD = 1 paragraph overview + 3 use cases. Complex auth system = full template. Don't generate filler for sections that don't apply.
+| # | Check | On failure → |
+|---|-------|-------------|
+| CC1 | Every story has at least 1 AS | Add missing AS |
+| CC2 | Every AS belongs to exactly 1 story | Assign orphan AS or delete |
+| CC3 | P0 stories have error path AS | Add error AS if missing |
+| CC4 | No 2 AS test the same behavior | Merge or delete duplicate |
+| CC5 | Constraints have AS verifying them | Add AS for uncovered constraints |
+| CC6 | Story count ≤7, AS count ≤20 | Go back to Phase 1 and split |
 
-Show the draft to the user. Wait for confirmation before generating the test plan.
+All checks must pass before showing the draft to the user.
+
+Show the draft to the user. Wait for confirmation before proceeding.
 
 ---
 
-## Phase 2: Clarify Ambiguities
+## Phase 3: Clarify Ambiguities
 
-Before generating the test plan, scan the spec for gaps. A test plan built on a vague spec produces vague tests.
+Before finalizing, scan the spec for gaps. Check BOTH the spec content AND the acceptance scenarios:
 
 | Lens | What to look for |
 |------|-----------------|
-| **Behavioral gaps** | Missing user actions, undefined system responses, incomplete flows |
-| **Data & persistence** | Undefined entities, missing relationships, unclear storage/lifecycle |
-| **Auth & access** | Who can do what is unclear, missing role definitions |
-| **Non-functional** | Vague adjectives without metrics ("fast", "secure", "scalable") — add SC-NNN with numbers |
-| **Integration** | Third-party API assumptions, unstated dependencies, SLA gaps |
-| **Concurrency & edge cases** | Multi-user scenarios, boundary conditions, error paths not addressed |
+| Behavioral gaps | Missing user actions, undefined system responses, incomplete flows. Which stories lack an error path AS? |
+| Data & persistence | Undefined entities, missing relationships, unclear storage/lifecycle |
+| Auth & access | Who can do what is unclear, missing role definitions |
+| Non-functional | Vague adjectives without metrics ("fast", "secure", "scalable") — add SC-NNN with concrete numbers |
+| Integration | Third-party API assumptions, unstated dependencies, SLA gaps |
+| Concurrency & edge cases | Multi-user scenarios, boundary conditions, error paths not addressed |
+| AS completeness | Which AS is missing Given or Then? |
+| AS overlap | Do 2 AS test the same behavior? |
+| Story orphans | Which story has no AS? |
+| Priority consistency | P0 story with only 1 happy path AS? |
+| Constraint coverage | Which constraint has no AS verifying it? |
 
 Identify the top 3-5 ambiguities (most impactful first). For each, ask the user a targeted question with 2-4 concrete options and a recommendation.
 
-If the spec is clear and complete, 0 questions is valid. Don't manufacture ambiguity.
+If 0 questions remain, you MUST state why — not just "spec is clear." Cite at minimum:
+- **Edge cases checked:** which boundary conditions were considered and found covered.
+- **Error paths checked:** which failure modes were verified to have AS.
+- **Integration points checked:** which external dependencies were reviewed.
+- One-line verdict per lens from the table above that had no findings.
+
+Example: *"0 questions. Edge cases: cart-empty and cart-max covered by AS-003/AS-004. Error paths: payment failure covered by AS-006. Auth: single-role feature, no ambiguity. No third-party integrations."*
+
+Don't manufacture ambiguity — but don't skip the justification either.
+
+**Present all questions (or the 0-question justification) to the user. Wait for answers before continuing.**
 
 Write clarifications back into the spec under `## Clarifications — <date>`.
-Then proceed to test plan generation.
+Update any affected stories or AS to reflect the user's answers.
+Then proceed to summary.
 
 ---
 
-## Phase 3: Generate the Test Plan
+## Phase 4: Summary
 
-Read the spec. For each section, extract:
-1. Use cases → at least 1 test (happy path) + 1 test (error path) each
-2. State transitions → test valid AND invalid transitions
-3. Constraints → test they hold under edge conditions
-4. Settings → test default AND non-default values
-5. Cross-cutting concerns (auth, validation) → integration-level tests
+Show:
+- Story counts (P0/P1/P2)
+- AS count
+- Directory structure created
+- Implementation order: which stories to implement first (by priority + dependency)
+- Next steps: "Implement stories in order. Use `/mf-test` to verify each story. For complex specs, run `/mf-challenge` first."
 
-Prioritize by risk: data loss/security = P0, error handling = P1, cosmetic/rare = P2.
+---
 
-### Output
+## Mode C: Update Flow
 
-Write to `docs/test-plans/<feature-name>.md`:
+> **⛔ CRITICAL — MANDATORY ORDER:**
+> Snapshot MUST be created **BEFORE** updating the spec.
+> If you update the spec first then create a snapshot → the snapshot contains the new content, old version is lost.
+> Correct order: C2 (classify) → C3 (snapshot) → C4 (report) → C5 (apply changes).
+> NEVER reverse the order of C3 and C5.
 
-```markdown
-# Test Plan: <Feature Name>
+### C0: Read current state
+
+Read `<feature>.md`. This is the current truth.
+
+### C1: Identify changes
+
+Compare the requested changes against the current spec. List:
+- Stories: added / modified / removed / unchanged
+- AS: added / modified / removed / unchanged
+- Constraints: added / modified / removed / unchanged
+
+### C2: Classification
+
+Walk through table M1-M6. If ANY condition is true → Major.
+
+| # | Condition | Example |
+|---|-----------|---------|
+| M1 | New story added | Adding S-004: Subscription |
+| M2 | Story removed | Removing S-002: Invoice |
+| M3 | Story priority changed | S-002 from P1 → P0 |
+| M4 | Story's main flow changed (Given or When changed) | AS-003 Given changes state, or When changes action |
+| M5 | Expected behavior changed (Then changed) for a P0 story | AS-001 Then changes result |
+| M6 | Constraint/invariant added or removed | Adding "balance must not be negative" |
+
+Minor = NONE of M1-M6 apply. Examples: typo fix, rewording without meaning change, adding/editing Data fields, formatting, adding Source ref.
 
-**Spec:** docs/specs/<feature-name>.md
-**Generated:** <$(date +%Y-%m-%d)>
+**Major → create snapshot before updating.**
+**Minor → no snapshot. Update directly.**
 
-## Test Cases
+> **⛔ MUST check ALL 6 conditions M1-M6.** Do not stop early.
+> Common mistake: check M1 = false, M2 = false → conclude Minor without checking M3-M6.
+> Correct: walk through M1 to M6 completely. If ANY is true → Major.
 
-| ID | Priority | Type | UC | FR/SC | Description | Expected |
-|----|----------|------|----|-------|-------------|----------|
-| TC-001 | P0 | unit | UC-001 | FR-001 | Valid login returns token | 200 + JWT |
-| TC-002 | P0 | unit | UC-001 | FR-002 | Wrong password returns 401 | 401 + error msg |
+### C3: Snapshot (if Major)
 
-## Implementation Order
-1. TC-001, TC-002 (no dependencies — start here)
-2. TC-003+ (depend on setup from earlier tests)
+If Major → create snapshot:
 
-## Coverage Notes
-- Highest risk areas: ...
-- Existing code needing modification: [file paths]
+**Step 1:** Copy file using shell command (bit-perfect, not through LLM):
+
+```bash
+mkdir -p docs/specs/<feature>/snapshots
+cp docs/specs/<feature>/<feature>.md docs/specs/<feature>/snapshots/<YYYY-MM-DD>.md
 ```
 
-**Priority:** P0 = must have (blocks release), P1 = should have, P2 = nice to have.
-**Type:** `unit`, `integration`, `e2e`, `snapshot`, `performance`
+If ref available: `cp ... snapshots/<YYYY-MM-DD>-<REF>.md`
+If same-day snapshot exists: `cp ... snapshots/<YYYY-MM-DD>-2.md`
+
+**Step 2:** Prepend header to the snapshot file (using Edit):
 
-### What NOT to produce
-- "Test that the feature works" — too vague
-- 50+ test cases for simple CRUD — over-testing
-- Testing implementation details — brittle
-- Duplicate tests verifying same behavior
+```markdown
+# Snapshot: <Feature Name>
+**Date:** <YYYY-MM-DD>
+**Ref:** <ticket/issue if available, "--" otherwise>
+**Reason:** <M1|M2|M3|M4|M5|M6 — list which conditions triggered>
 
 ---
 
-## Phase 4: Summary
+```
+
+Header is added BEFORE the copied content. Do not modify any other content in the snapshot.
+
+> **⛔ Why `cp` instead of LLM copy:** Specs require 101% accuracy. LLM text copy risks
+> dropping lines, altering formatting, truncating long content. `cp` is bit-perfect.
 
-Show: test case counts (P0/P1/P2), implementation order, estimated scope.
-Next steps: "Use `/mf-test` after each chunk. For complex plans, run `/mf-challenge` first."
+**Step 3:** Rotate snapshots. Check the spec frontmatter for `Snapshot limit: N`. If absent, default to **5**.
+After creating a new snapshot, if `snapshots/` contains more files than the limit:
+- Sort by timestamp in filename.
+- Delete oldest files until count equals the limit.
+- Only delete snapshot files. Log deletion in Change Log: `"Snapshot <filename> rotated out"`.
+
+If Minor: skip C3 entirely.
+
+**Snapshots are immutable.** Never edit a created snapshot. Wrong snapshot → create a new one, delete the wrong one.
+
+**mf-plan creates snapshots. Developers do not create them manually.** Developers do not decide, intervene, or skip.
+
+### C4: Change report
+
+Display to the user:
+
+```markdown
+## Change Report: <feature>
+**Classification:** Major / Minor
+**Snapshot:** Created <filename> / Not needed
+
+### Changes
+| Item | Action | Detail |
+|------|--------|--------|
+| S-002 | Priority change | P1 → P0 |
+| AS-003 | Updated | Then changed |
+| S-004 | Added | Subscription (P1) |
+
+### Unchanged
+S-001, S-003
+```
+
+> **⛔ MUST wait for user confirmation before applying.**
+> Do not show the report and apply in the same step.
+> User has the right to reject or modify the change report.
+
+### C5: Apply changes
+
+- Update the spec directly.
+- Update `Last updated`.
+- Write to Change Log.
+- New AS use the next sequential ID (never reuse deleted IDs).
+- New AS follow the same Writing Instructions as Phase 2 (concrete values, one behavior per AS, no vague/duplicate/implementation-testing scenarios).
+
+> **⛔ Change Log MUST be updated at this step.**
+> Common mistake: update the spec, forget to write to Change Log.
+> Every C5 execution → Change Log MUST have a new row. No exceptions.
+> (Exception: non-semantic changes — C7 — do not write to Change Log.)
+
+### C6: Consistency check
+
+After updating, verify:
+
+| # | Check | On failure → |
+|---|-------|-------------|
+| CC1 | Every story has at least 1 AS | Add missing AS |
+| CC2 | Every AS belongs to exactly 1 story | Assign orphan AS or delete |
+| CC3 | P0 stories have error path AS | Add error AS if missing |
+| CC4 | No 2 AS test the same behavior | Suggest merge or delete duplicate |
+| CC5 | Constraints have AS verifying them | Add AS for uncovered constraints |
+| CC6 | Story count ≤7, AS count ≤20 | Suggest splitting spec (Phase 1) |
+
+> **⛔ Consistency check is NOT optional.**
+> Run CC1-CC6 after EVERY update (Major and Minor).
+> Common mistake: finish update, looks fine → skip consistency check.
+> CC6 (size check) is especially easy to skip — MUST check after every story/AS addition.
+
+If any check fails → fix or report to user. NEVER skip.
+
+### C7: Non-semantic changes
+
+If the change is only typo, formatting, or wording that does NOT change behavior:
+- Edit directly, do not run C2-C6.
+- Do not write to Change Log.
+- Do not create snapshot.
+
+Criteria for "non-semantic": Given, When, Then, priority, constraint **DO NOT** change in meaning.
+
+> **⛔ When in doubt whether "non-semantic or behavioral?" → treat as behavioral.**
+> Common mistake: LLM classifies a Then change as "rewording" to avoid snapshot overhead.
+> Test: if a developer reads the AS before and after the change and would write different code → it is behavioral.
+
+### C8: Archival (all stories removed)
+
+If a Mode C update results in ALL stories being removed from a spec:
+
+1. Create a snapshot per C3 (this is a Major change — M2 applies).
+2. Move the entire feature directory to `docs/specs/_archived/`:
+   ```bash
+   mkdir -p docs/specs/_archived
+   mv docs/specs/<feature> docs/specs/_archived/$(date +%Y-%m-%d)-<feature>
+   ```
+3. The archived directory retains all snapshots and the final spec state.
+4. Log in Change Log before archiving: `"Feature archived — all stories removed"`.
+
+Archived specs are read-only. To resurrect a feature, copy from `_archived/` back to `docs/specs/` and run `/mf-plan` in Mode A.
+
+---
 
 ## Naming Convention
 
-Spec and test plan MUST share the same filename:
 ```
-docs/specs/<feature-name>.md       ← kebab-case, 2-3 words
-docs/test-plans/<feature-name>.md  ← same name
+docs/specs/<feature>/              ← kebab-case, 2-3 words
+  <feature>.md                     ← same name as directory
+  snapshots/
+    YYYY-MM-DD.md
+    YYYY-MM-DD-<REF>.md
 ```
-- Use feature name, not module name: `user-auth.md` not `AuthService.md`
+
+- Feature name, not module name: `user-auth/` not `AuthService/`
+- Sub-specs when splitting: `<feature>-<sub>.md` in the same directory
 - No prefix/suffix: `user-auth.md` not `spec-user-auth.md`
 
-**Requirement IDs** — sequential per spec:
-- `UC-001` Use Case, `FR-001` Functional Requirement, `SC-001` Success Criteria, `TC-001` Test Case
-- Every TC must reference at least one FR or SC for traceability.
+**ID rules:**
+- `S-NNN` Story — sequential per spec, starting from S-001
+- `AS-NNN` Acceptance Scenario — sequential per spec, across all stories, starting from AS-001
+- `FR-NNN` Functional Requirement — if needed
+- `SC-NNN` Success Criteria — if needed
+- Deleted IDs must never be reused
+- **Sub-spec numbering is local.** Each sub-spec starts its own S-001, AS-001 sequence.
+  Sub-specs are self-contained (Phase 1 rule), so IDs need not be globally unique.
+- **Cross-references between sub-specs** use the sub-spec name as prefix:
+  `billing-refund:AS-002` refers to AS-002 in `billing-refund.md`.
+  Avoid cross-references where possible — if you need many, the split may be wrong.
+
+---
 
 ## Rules
-1. **Spec-first.** Test plan derives from spec, never from code.
-2. **Codebase-aware.** Don't plan features that already exist.
-3. **Actionable.** Every test case must be unambiguous enough to implement directly.
-4. **Proportional.** Simple feature = simple plan. Don't over-engineer CRUD.
-5. **Traceable.** Every test links to a use case. No orphan tests.
-6. **Consistent names.** Spec and test plan always share the same filename.
+
+1. **Spec-first.** Code serves the spec, not the other way around.
+2. **Single file = current truth.** `<feature>.md` always reflects the current state.
+3. **Codebase-aware.** Don't plan features that already exist.
+4. **Actionable.** Every AS must be clear enough to implement directly.
+5. **Proportional.** Simple feature = simple spec. Don't over-engineer CRUD.
+6. **Traceable.** Every AS belongs to 1 story. No orphan AS.
+7. **Bounded.** Spec exceeding 7 stories or 20 AS must be split.
+8. **Snapshot = mf-plan's job.** Developers do not create, delete, or edit snapshots.
+9. **Classification = checklist.** Major/Minor decided by table M1-M6, not judgment.
+10. **ID immutable.** Assigned IDs never change, never get reused.
+
+---
+
+## Traps — Common Mistakes That MUST Be Avoided
+
+| # | Trap | Consequence | Rule violated |
+|---|------|-------------|--------------|
+| TRAP-1 | Update spec BEFORE creating snapshot | Snapshot contains new content, old version lost | Order C3→C5 |
+| TRAP-2 | Check M1-M2 then stop, skip M3-M6 | Major change classified as Minor, snapshot missed | Classification M1-M6 |
+| TRAP-3 | Skip consistency check (CC1-CC6) | Story without AS, P0 missing error path, spec bloat undetected | C6 Consistency |
+| TRAP-4 | Classify behavioral change as "non-semantic" | Important change not snapshotted, not logged | C7 Non-semantic |
+| TRAP-5 | Apply changes without waiting for user confirmation | User loses control, wrong changes can't be rolled back | C4 Change report |
+| TRAP-6 | Update spec but forget to write Change Log | Change history lost, no one knows what happened | C5 Apply |
+| TRAP-7 | Reuse deleted ID (S-003 deleted then assigned to new story) | Confusion with old references in code, commits, conversations | ID rules |
+| TRAP-8 | LLM copies spec content instead of using `cp` for snapshot | Lines dropped, formatting altered, truncation — inaccurate snapshot | C3 Snapshot |
+| TRAP-9 | Skip Phase 1 (Scope & Split) for large features | Spec bloats >7 stories, hard to maintain, hard to review | Phase 1 |
+| TRAP-10 | Write P2-depth AS for a P0 story | P0 story lacks Given/When/Then/Data, developer can't implement | AS Depth |

package/templates/.claude/commands/mf-review.md

@@ -7,7 +7,7 @@ Pre-merge code review — security, correctness, spec alignment.
    BASE=$(git symbolic-ref refs/remotes/origin/HEAD 2>/dev/null | sed 's|refs/remotes/origin/||') || BASE="main"
    git log --oneline "$BASE"...HEAD
    ```
-2. Check for spec in `docs/specs/` and test plan in `docs/test-plans/` — review against INTENT.
+2. Check for spec in `docs/specs/<feature>/<feature>.md` — review against INTENT.
 3. Read the diff: `git diff "$BASE"...HEAD`
 
 If `$ARGUMENTS` provided → scope to those files only.
@@ -50,11 +50,15 @@ Spend 60% of analysis on the primary focus. Cover all categories, but proportion
 - **Null safety:** Optionals used without guards? `object!.property` without nil check?
 
 ### Spec-Test Alignment (Medium)
-- Source changed but no spec update in `docs/specs/`? → flag
+- Source changed but no spec update in `docs/specs/<feature>/`? → flag
 - Source changed but no test update? → flag
-- Spec changed but tests not updated? → flag
+- Spec changed but acceptance scenarios or tests not updated? → flag
 - Code removed but dead tests remain? → flag
 - Spec contains vague requirements without metrics ("fast", "secure", "easy", "scalable")? → flag with suggestion to add SC-NNN with concrete numbers
+- **AS-to-test name check:** Read the spec's `## Stories` section. For each AS-NNN, check if a test file contains a test named or described with that AS ID or its short description. Flag:
+  - AS in spec with no matching test → "AS-NNN: \<description\> has no corresponding test"
+  - Test referencing an AS-NNN that no longer exists in the spec → "Test references removed AS-NNN"
+  Keep this lightweight — match on AS-NNN identifiers and story name substrings, not semantic analysis.
 
 ### Code Quality (Medium)
 - Dead code: removed functions still imported elsewhere?

package/templates/.claude/commands/mf-test.md

@@ -1,4 +1,4 @@
-Write tests from test plan, compile, run, fix until green.
+Write tests from spec acceptance scenarios, compile, run, fix until green.
 
 ## Phase 0: Build Context
 
@@ -10,8 +10,7 @@ Write tests from test plan, compile, run, fix until green.
    If `$ARGUMENTS` provided → scope to that file or feature only.
    If no changes → "No source changes found. Specify a file or feature."
 
-2. **Read the test plan** in `docs/test-plans/` if it exists — this is your roadmap.
-3. **Read the spec** in `docs/specs/` if it exists — understand the INTENT behind the code.
+2. **Read the spec** at `docs/specs/<feature>/<feature>.md` — the `## Stories` section with acceptance scenarios is your roadmap. The `## Overview` and `## Constraints` sections tell you the INTENT behind the code.
 4. **Read existing tests** for the changed files — find patterns, fixtures, naming conventions. Don't duplicate.
 
 ---
@@ -86,14 +85,25 @@ If tests fail:
 
 ```
 Tests: X added, Y modified, Z unchanged
-Result: All passing ✓
+Result: All passing ✓ / N failing ✗
 Coverage: [critical uncovered paths if any]
 Files: [test files touched]
-Plan: [TC-001 ✓, TC-002 ✓, TC-005 new]
+Stories: [AS-001 ✓, AS-002 ✓, AS-005 new]
 ```
 
-If behavior changed: "Consider updating the spec in docs/specs/."
+If behavior changed: "Consider updating the spec in docs/specs/<feature>/<feature>.md."
+
+### Spec Gap Detection
+
+If a test fails due to an edge case, error path, or boundary condition that is NOT covered by any existing AS in the spec:
+
+1. State explicitly: **"This failure suggests a missing acceptance scenario."**
+2. Describe the gap: what behavior was tested, which story it belongs to, why no AS covers it.
+3. Prompt: **"Run `/mf-plan <spec-path> 'Add AS for <description>'` to add the missing scenario."**
+
+Do not silently fix the test and move on. A test that has no corresponding AS means the spec is incomplete — the spec must be updated first.
 
 ## Rules
 1. **Behavior over implementation.** Test what code DOES, not how.
 2. **Independent tests.** Each test sets up its own state, cleans up after.
+3. **Spec stays upstream.** If a test reveals a spec gap, update the spec before adding the test.