dev-playbooks 1.5.1 → 1.5.3

package/README.md

@@ -81,11 +81,21 @@ npx dev-playbooks@latest init
 
 ### Install targets
 
-After initialization:
-- Claude Code: `~/.claude/skills/devbooks-*`
-- Codex CLI: `~/.codex/skills/devbooks-*`
-- Qoder: `~/.qoder/` (manual setup required)
-- OpenCode: `~/.config/opencode/skill/devbooks-*`
+During initialization, you can choose the Skills installation scope:
+
+| Scope | Description | Path Example |
+|-------|-------------|--------------|
+| **Project** (default) | Available only in this project | `.claude/skills/devbooks-*` |
+| **Global** | Shared across all projects | `~/.claude/skills/devbooks-*` |
+
+```bash
+# Interactive selection
+dev-playbooks init
+
+# Non-interactive
+dev-playbooks init --tools claude --scope project  # Project-level
+dev-playbooks init --tools claude --scope global   # Global
+```
 
 ### Quick integration
 
@@ -194,6 +204,7 @@ Run devbooks-archiver skill for change add-oauth2
 |-------|-------------|
 | devbooks-entropy-monitor | System entropy metrics |
 | devbooks-brownfield-bootstrap | Brownfield project bootstrap |
+| devbooks-convergence-audit | Convergence audit (anti-deception design) |
 
 ---
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "dev-playbooks",
-  "version": "1.5.1",
+  "version": "1.5.3",
   "description": "AI-powered spec-driven development workflow",
   "keywords": [
     "devbooks",

package/skills/Skills-Usage-Guide.md

@@ -344,6 +344,54 @@ If you are not using DevBooks, replace `dev-playbooks/specs` / `dev-playbooks/ch
 
 ---
 
+## `devbooks-test-reviewer` (Test Reviewer) [New]
+
+- Purpose: review test quality in `tests/` (coverage, boundary conditions, readability, maintainability); output review comments only; do not modify code.
+- When to use:
+  - After Test Owner completes tests, an independent quality review is needed
+  - You want to find coverage gaps, missing boundary conditions, and maintainability issues
+- Copy-paste prompt:
+  ```text
+  You are Test Reviewer. Explicitly use `devbooks-test-reviewer`.
+  Review test quality only (coverage/boundary conditions/readability/maintainability); do not modify code.
+  Inputs: `tests/**` + `dev-playbooks/changes/<change-id>/verification.md` (if present)
+  Output: coverage gaps / missing boundary conditions / maintainability risks / improvement suggestions.
+  ```
+
+---
+
+## `devbooks-convergence-audit` (Convergence Auditor) [New]
+
+- Purpose: evaluate DevBooks workflow convergence using "evidence-first, distrust declarations" principle; detect "Sisyphus anti-pattern" and "false completion". Actively verify rather than trust document assertions.
+- Anti-confusion design: never trust any assertion in documents (Status: Done, AC checked, etc.); must confirm through verifiable evidence.
+- When to use:
+  - You want to evaluate true workflow convergence (not document claims)
+  - You suspect a change package is "false completion" or has "stale evidence"
+  - You want to detect Sisyphus trap (repeated rework without convergence)
+- Copy-paste prompt:
+  ```text
+  You are Convergence Auditor. Explicitly use `devbooks-convergence-audit`.
+  First read: `dev-playbooks/project.md` (if present)
+  Goal: perform a convergence audit on a specified change package.
+
+  Execute with anti-confusion principles:
+  1) Read all documents under `dev-playbooks/changes/<change-id>/`
+  2) Verify each "declaration" (Status=Done, AC checked, etc.) against evidence
+  3) Actually run tests for verification (if possible)
+  4) Output: declaration vs evidence comparison table + trustworthiness score + confusion detection results
+
+  Change package: <change-id>
+  truth-root: dev-playbooks/specs
+  change-root: dev-playbooks/changes
+  ```
+- Trustworthiness scoring:
+  - 90-100: ✅ Trustworthy convergence, continue current workflow
+  - 70-89: ⚠️ Partially trustworthy, needs additional verification
+  - 50-69: 🟠 Questionable, parts need rework
+  - < 50: 🔴 Untrustworthy (Sisyphus trap), needs full review
+
+---
+
 ## Generating SCIP Index (Manual)
 
 If you need graph-based code understanding (call graph, impact analysis, symbol references), you can generate a SCIP index manually:

package/skills/devbooks-test-reviewer/SKILL.md

@@ -53,6 +53,34 @@ Before execution, you **must** search for configuration in the following order (
 - Must compare against `verification.md` to check if tests cover all ACs
 - If missing tests are found, clearly indicate the missing AC-ID
 
+### CON-ROLE-004: Only Review Test Code Quality, Not Runtime Results
+> Core principle: Test Reviewer focuses on "**how well tests are written**", not "**whether tests pass**"
+
+**Prohibited**:
+- Discussing whether tests Pass/Fail, Red/Green, or Skip
+- Commenting on "functionality not implemented" or "missing dependencies"
+- Providing implementation suggestions to Coder (e.g., "implement xxx command", "add xxx field")
+- Referencing workflow phases (e.g., "Red baseline phase", "this is expected behavior")
+- Judging test quality based on test runtime results
+
+**Allowed**:
+- Reviewing correctness and clarity of test assertions
+- Reviewing test structure and organization
+- Reviewing completeness of test scenarios (based on code analysis, not runtime results)
+
+### CON-ROLE-005: Coverage = Test Existence, Not Test Pass Status
+
+| Coverage Status | Definition | Judgment Criteria |
+|-----------------|------------|-------------------|
+| ✅ Covered | Corresponding test case exists for AC | Test/it block exists in test file |
+| ⚠️ Partially Covered | Test exists but scenarios incomplete | Missing boundary conditions, error paths, etc. |
+| ❌ Missing | No corresponding test case | No matching test code found |
+
+**Prohibited** from including the following in coverage assessment:
+- Test Skip status (skipped tests still count as "Covered")
+- Test Fail status (failing tests still count as "Covered")
+- Test runtime or performance
+
 ---
 
 ## Review Dimensions
@@ -104,11 +132,13 @@ Before execution, you **must** search for configuration in the following order (
 
 ## Coverage Analysis
 
+> **Note**: Coverage status is based on test code existence, not test runtime results (Pass/Fail/Skip).
+
 | AC-ID | Test File | Coverage Status | Notes |
 |-------|-----------|-----------------|-------|
-| AC-001 | test-a.ts | Covered | - |
-| AC-002 | - | Missing | Needs to be added |
-| AC-003 | test-b.ts | Partially Covered | Missing boundary conditions |
+| AC-001 | test-a.ts | ✅ Covered | Test case exists |
+| AC-002 | - | ❌ Missing | No corresponding test code |
+| AC-003 | test-b.ts | ⚠️ Partially Covered | Test exists but missing boundary conditions |
 
 ## Issue List
 
@@ -178,7 +208,7 @@ After completing the review, you **must** provide a clear conclusion:
 | Skill Name | devbooks-test-reviewer |
 | Phase | Apply |
 | Artifact | Review report (not written to change package) |
-| Constraints | CON-ROLE-001~003 |
+| Constraints | CON-ROLE-001~005 |
 
 ---