bmad-module-skill-forge 0.6.0 → 0.7.1

package/src/workflows/test-skill/validation-report.md

@@ -1,390 +0,0 @@
----
-validationDate: 2026-02-26
-completionDate: 2026-02-26
-workflowName: test-skill
-workflowPath: _bmad-output/bmb-creations/workflows/test-skill
-validationStatus: COMPLETE
----
-
-# Validation Report: test-skill
-
-**Validation Started:** 2026-02-26
-**Validation Completed:** 2026-02-26
-**Validator:** BMAD Workflow Validation System
-**Standards Version:** BMAD Workflow Standards
-
----
-
-## File Structure & Size
-
-### Folder Structure: PASS
-
-```
-test-skill/
-├── workflow.md                         (57 lines)
-├── data/
-│   ├── scoring-rules.md               (72 lines)
-│   └── output-section-formats.md      (97 lines)
-├── steps-c/
-│   ├── step-01-init.md                 (185 lines) ✅
-│   ├── step-02-detect-mode.md          (162 lines) ✅
-│   ├── step-03-coverage-check.md       (222 lines) ⚠️ Approaching limit
-│   ├── step-04-coherence-check.md      (249 lines) ⚠️ Approaching limit
-│   ├── step-05-score.md                (222 lines) ⚠️ Approaching limit
-│   └── step-06-report.md              (175 lines) ✅
-└── templates/
-    └── test-report-template.md         (24 lines)
-```
-
-### Required Files: PASS
-
-- ✅ workflow.md exists
-- ✅ All 6 design steps have corresponding files
-- ✅ Sequential numbering (01-06) with no gaps
-- ✅ Final step (step-06) exists
-- ✅ Data folder with scoring rules and output formats
-- ✅ Templates folder with output template
-
-### File Size Analysis: PASS (after remediation)
-
-| File | Lines | Status |
-|------|-------|--------|
-| step-01-init.md | 185 | ✅ Good |
-| step-02-detect-mode.md | 162 | ✅ Good |
-| step-03-coverage-check.md | 222 | ⚠️ Approaching limit |
-| step-04-coherence-check.md | 249 | ⚠️ Approaching limit (fixed from 294) |
-| step-05-score.md | 222 | ⚠️ Approaching limit |
-| step-06-report.md | 175 | ✅ Good (fixed from 254) |
-
-**Remediation applied:**
-- Created `data/output-section-formats.md` — extracted output templates from step-04 and step-06
-- step-04: 294 → 249 lines (output format templates moved to data file)
-- step-06: 254 → 175 lines (gap format templates and remediation examples moved to data file)
-
----
-
-## Frontmatter Validation
-
-### Frontmatter Variables: PASS
-
-All 21 frontmatter variables across 7 files are consumed in their step bodies.
-
-| Step | Variables | All Used | Status |
-|------|-----------|----------|--------|
-| workflow.md | name, description, web_bundle, installed_path | Yes (metadata fields) | PASS |
-| step-01 | nextStepFile, outputFile, templateFile, sidecarFile, skillsOutputFolder | Yes | PASS |
-| step-02 | nextStepFile, outputFile | Yes | PASS |
-| step-03 | nextStepFile, outputFile, scoringRulesFile | Yes | PASS |
-| step-04 | nextStepFile, outputFile, outputFormatsFile | Yes | PASS |
-| step-05 | nextStepFile, outputFile, scoringRulesFile | Yes | PASS |
-| step-06 | outputFile, scoringRulesFile, outputFormatsFile | Yes | PASS |
-
-### Path Format: PASS
-
-- Step-to-step paths use `./filename.md` format (5/5)
-- Parent-folder paths use `../filename.md` format (6/6)
-- External paths use `{project-root}` prefix (2/2)
-- No `{workflow_path}` variable present (7/7)
-- All files have `name` + `description` fields (7/7)
-
----
-
-## Critical Path Violations
-
-### Hardcoded Paths in Step Bodies: PASS
-
-All 6 step files scanned — zero violations. All path references use frontmatter variables.
-
-### Workflow.md Bootstrap Paths: PASS (with 2 standard warnings)
-
-| Line | Path | Context | Verdict |
-|------|------|---------|---------|
-| 50 | `{project-root}/_bmad/skf/config.yaml` | Config bootstrap | WARNING — standard entrypoint pattern |
-| 57 | `./steps-c/step-01-init.md` | First step bootstrap | WARNING — standard entrypoint pattern |
-
-Both are expected bootstrap patterns for workflow entrypoints.
-
-### File Reference Chain: PASS
-
-| Source | Reference | Target Exists |
-|--------|-----------|---------------|
-| step-01 → step-02 | `./step-02-detect-mode.md` | ✅ |
-| step-02 → step-03 | `./step-03-coverage-check.md` | ✅ |
-| step-03 → step-04 | `./step-04-coherence-check.md` | ✅ |
-| step-04 → step-05 | `./step-05-score.md` | ✅ |
-| step-05 → step-06 | `./step-06-report.md` | ✅ |
-| step-06 | (final step — no nextStepFile) | ✅ |
-| step-01 → template | `../templates/test-report-template.md` | ✅ |
-| step-03, 05, 06 → scoring | `../data/scoring-rules.md` | ✅ |
-| step-04, 06 → formats | `../data/output-section-formats.md` | ✅ |
-
----
-
-## Menu Handling Validation
-
-### Menu Compliance: PASS (71/71 checks)
-
-| Step | Expected Pattern | Actual | Status |
-|------|-----------------|--------|--------|
-| step-01 | Auto-proceed (init) | Auto-proceed, no user menu | PASS |
-| step-02 | Auto-proceed | Auto-proceed, no user menu | PASS |
-| step-03 | Auto-proceed | Auto-proceed, no user menu | PASS |
-| step-04 | Auto-proceed | Auto-proceed, no user menu | PASS |
-| step-05 | Auto-proceed | Auto-proceed, no user menu | PASS |
-| step-06 | C-only (final) | `[C] Finish` with halt-and-wait | PASS |
-
-- ✅ No A/P options in any auto-proceed step
-- ✅ Step-06 has "halt and wait" execution rule
-- ✅ Step-06 redisplays menu for non-C input
-- ✅ Step-06 correctly omits "loading next step" universal rule
-
----
-
-## Step Type Validation
-
-### Step Type Compliance: PASS (71/71 checks)
-
-| Step | Expected Type | Match | All Sections Present | Status |
-|------|--------------|-------|---------------------|--------|
-| step-01 | Init + Input Discovery | ✅ | 6/6 | PASS |
-| step-02 | Middle (Simple) | ✅ | 6/6 | PASS |
-| step-03 | Validation Sequence | ✅ | 6/6 | PASS |
-| step-04 | Validation Sequence (conditional) | ✅ | 6/6 | PASS |
-| step-05 | Middle (Simple) | ✅ | 6/6 | PASS |
-| step-06 | Final Step (C-only) | ✅ | 6/6 | PASS |
-
-All steps have: STEP GOAL, MANDATORY EXECUTION RULES, EXECUTION PROTOCOLS, CONTEXT BOUNDARIES, MANDATORY SEQUENCE, SUCCESS/FAILURE METRICS.
-
-**Remediation applied:** Fixed step-06 section numbering gap (was 1,2,3,4,6,7,8 → now 1,2,3,4,5,6,7).
-
----
-
-## Output Format Validation
-
-### Template Type: PASS
-
-Template matches "structured" type with:
-- YAML frontmatter with defined metadata fields
-- Clear section headers: Test Summary, Coverage Analysis, Coherence Analysis, Completeness Score, Gap Report
-- Skeletal structure (headers only) — each step fills its section
-
-### Step-to-Output Mapping: PASS
-
-| Step | Action | Output Section | outputFile in FM | Appends Before Next |
-|------|--------|---------------|------------------|-------------------|
-| 01 | Creates doc | (from template) | ✅ | ✅ |
-| 02 | Appends | Test Summary | ✅ | ✅ |
-| 03 | Appends | Coverage Analysis | ✅ | ✅ |
-| 04 | Appends | Coherence Analysis | ✅ | ✅ |
-| 05 | Appends | Completeness Score | ✅ | ✅ |
-| 06 | Appends | Gap Report | ✅ | ✅ (final step) |
-
-Every step has a CRITICAL STEP COMPLETION NOTE enforcing write-before-proceed.
-
-### Final Polish Step: PASS (N/A)
-
-Not required — structured template does not need editorial polish.
-
----
-
-## Validation Design Check
-
-### Validation Steps Required? N/A — PASS
-
-test-skill IS itself a testing/validation workflow. It is the quality gate that tests skills. Adding validation steps (`steps-v/`) to a validation workflow would be circular.
-
-- ✅ No `steps-v/` folder — correct for workflow purpose
-- ✅ Workflow plan confirms: "Lifecycle Support: create-only" (no steps-e/ or steps-v/)
-- ✅ The workflow's entire purpose is testing — every step IS a validation activity
-
----
-
-## Instruction Style Check
-
-### Domain Assessment: Prescriptive (correct)
-
-test-skill is an audit/testing workflow (Ferris Audit mode) with deterministic analysis procedures, defined formulas, and exact comparison logic. Prescriptive style is appropriate.
-
-### Per-Step Classification: PASS
-
-| Step | Classification | Appropriate | Status |
-|------|---------------|-------------|--------|
-| step-01 | Prescriptive | ✅ Yes | PASS |
-| step-02 | Prescriptive | ✅ Yes | PASS (1 warning) |
-| step-03 | Prescriptive | ✅ Yes | PASS |
-| step-04 | Prescriptive | ✅ Yes | PASS (1 warning) |
-| step-05 | Prescriptive | ✅ Yes | PASS |
-| step-06 | Prescriptive | ✅ Yes | PASS |
-
-**Warnings:**
-- ⚠️ step-02 lines 62-66: Skill type indicators mix definitive metadata checks with interpretive signals ("Self-contained API surface"). These are supplementary confirmation signals, not primary decision criteria (metadata field is checked first).
-- ⚠️ step-04 lines 160-165: Integration pattern completeness criteria use qualitative language ("complete flow, not fragments"). Could benefit from more concrete criteria.
-
-**Remediation applied:** Added coherence score aggregation formula to `scoring-rules.md` (was missing — agents would have had to guess how to combine reference validity and integration completeness).
-
----
-
-## Collaborative Experience Check
-
-### Interaction Pattern: Mostly Autonomous — PASS
-
-| Step | Interaction | Quality |
-|------|------------|---------|
-| step-01 | Input discovery | Natural, progressive, single question — not a laundry list |
-| step-02 | Autonomous | Clean auto-proceed, informative mode report |
-| step-03 | Autonomous | Good progress reporting with coverage metrics |
-| step-04 | Autonomous | Clean naive/contextual branching with proper fallbacks |
-| step-05 | Autonomous | Excellent mathematical transparency ("show the math") |
-| step-06 | Terminal menu | Clear results presentation, appropriate C-only menu |
-
-### Workflow Progression: PASS
-
-```
-Init → Classify → Analyze Coverage → Analyze Coherence → Score → Report
-```
-
-Each step has explicit CONTEXT BOUNDARIES with FORBIDDEN markers preventing scope creep. Steps build logically on each other. Append-only document serves as state mechanism.
-
-### Error Handling: Good
-
-- step-01: Four-tier handling (hard stop for missing SKILL.md, warning for missing metadata.json, graceful default for missing forge-tier.yaml). Checks `preferences.yaml` for `tier_override` after loading forge tier.
-- step-03/04: Subprocess fallback rules at multiple locations
-- step-06: Zero-gaps clean pass path
-
-**Warning:** step-03 has no explicit error handling for missing/inaccessible source files at `source_path`. Minor edge case.
-
-### Overall Experience Quality: **Good**
-
----
-
-## Subprocess Optimization Opportunities
-
-### Usage Assessment: Good
-
-| Step | Patterns Used | Fallback | Status |
-|------|--------------|----------|--------|
-| 01 | None (not needed) | Universal rule | PASS |
-| 02 | None (not needed) | Universal rule | PASS |
-| 03 | Pattern 2 (per-file AST analysis) | 3 locations | PASS |
-| 04 | Pattern 1 (grep refs) + Pattern 2 (per-ref validation) | 3 locations | PASS |
-| 05 | None (not needed) | Universal rule | PASS |
-| 06 | None (not needed) | Universal rule | PASS |
-
-Subprocess optimization correctly concentrated in steps 03 and 04 (heavy analysis). Lighter steps correctly avoid unnecessary overhead. All steps have subprocess fallback rules.
-
-### Minor Opportunities:
-
-- ⚠️ step-03 Quick tier: Could use Pattern 1 (single grep) for pre-scan before per-file analysis
-- ⚠️ step-04: Per-reference validations are independent — could note Pattern 4 (parallel) opportunity
-
-### Data Files: Well-Structured
-
-- `scoring-rules.md` (72 lines): Clear tables, formulas, severity rules
-- `output-section-formats.md` (97 lines): Self-contained format templates per section
-
----
-
-## Cohesive Review
-
-### Overall Assessment: **Good**
-
-**Cohesiveness:** Excellent — all 6 steps build logically. Clean separation of concerns with each step having a single responsibility.
-
-**Data Flow:** Good — append-only document building with proper frontmatter state tracking. One minor gap: step 02 references `{threshold}%` in Test Summary before step 05 loads it from scoring rules.
-
-**Consistency:** Excellent — identical structure pattern across all steps (frontmatter → goal → rules → protocols → boundaries → sequence → metrics). Same role reinforcement ("Ferris Audit mode"), same emoji conventions, same subprocess fallback wording.
-
-**Goal Achievement:** Good — would produce a useful, structured test report with pass/fail determination and actionable gap remediation. Tier-aware depth scaling ensures graceful degradation.
-
-### Strengths:
-
-1. Rigorous traceability (zero hallucination with file:line citations)
-2. Four-tier depth scaling (Quick/Forge/Forge+/Deep)
-3. Clean separation of concerns with explicit FORBIDDEN markers
-4. Well-designed subprocess optimization with structured return formats
-5. Actionable remediation quality rules with good/bad examples
-6. Data files properly externalized for reuse
-
-### Weaknesses:
-
-1. Weight tables duplicated between `scoring-rules.md` and step-05 inline — maintenance risk
-2. Batch mode referenced in workflow plan but not implemented in step files
-3. Quick tier pass may give false confidence (name-matching only)
-
----
-
-## Plan Quality Validation
-
-### Plan Completeness: PASS
-
-All 7 required sections present and substantive. Status: BUILD_COMPLETE with all 9 pipeline steps completed.
-
-### Design-to-Implementation: PASS (1 warning fixed)
-
-- ✅ All 6 planned steps have corresponding files with matching names
-- ✅ Step descriptions align with plan goals
-- ✅ File structure matches plan (fixed: added `output-section-formats.md` to plan)
-
-### Consistency: PASS (1 warning)
-
-- ✅ Step types match plan
-- ✅ Menu patterns match plan
-- ✅ Data flow matches plan
-- ⚠️ Subprocess Pattern 4 (batch parallel) attributed to step 03 in plan table but is actually a workflow-level orchestration concern
-
-### Plan Quality: PASS
-
-- Conversion notes are insightful with genuine analysis
-- Classification decisions well-reasoned with structure implications
-- Requirements clear and complete with testable success criteria
-- Design specification detailed enough to rebuild the workflow
-
----
-
-## Summary
-
-### Validation Complete: 2026-02-26
-
-**Overall Status: PASS**
-
-### Results Overview
-
-| Validation Step | Result |
-|----------------|--------|
-| File Structure & Size | PASS (after remediation) |
-| Frontmatter Validation | PASS |
-| Critical Path Violations | PASS (2 standard warnings) |
-| Menu Handling Validation | PASS (71/71) |
-| Step Type Validation | PASS (71/71) |
-| Output Format Validation | PASS |
-| Validation Design Check | PASS (N/A) |
-| Instruction Style Check | PASS (2 warnings) |
-| Collaborative Experience Check | PASS |
-| Subprocess Optimization | Good |
-| Cohesive Review | Good |
-| Plan Quality Validation | PASS (1 warning fixed) |
-
-### Issues Summary
-
-**Critical Issues:** 0
-**Blocking Issues:** 0
-**Warnings (non-blocking):** 7
-
-1. step-02: Interpretive skill type indicators alongside definitive metadata check
-2. step-04: Qualitative integration completeness criteria
-3. step-03: No error handling for missing source files at source_path
-4. step-05: Weight tables duplicated from scoring-rules.md
-5. step-02: Threshold referenced before loaded by step-05
-6. step-04: Could note Pattern 4 (parallel) for independent reference validations
-7. Batch mode in plan but not in step files (unimplemented feature)
-
-### Remediation Applied During Validation
-
-1. Created `data/output-section-formats.md` — extracted templates to reduce step-04 and step-06 sizes
-2. Fixed step-06 section numbering gap (4 → 6 became 4 → 5)
-3. Added coherence score aggregation formula to `scoring-rules.md`
-4. Updated plan file structure to include `output-section-formats.md`
-
-### Recommendation
-
-**Ready to use.** The test-skill workflow is well-architected with strong traceability, clean separation of concerns, and proper subprocess optimization. The 7 non-blocking warnings are minor quality improvements that can be addressed in a future iteration. The workflow will function correctly for single-skill testing in all three forge tiers.

package/src/workflows/test-skill/workflow-plan-test-skill.md

@@ -1,292 +0,0 @@
----
-conversionFrom: 'src/workflows/test-skill/test-skill.spec.md'
-originalFormat: 'BMAD Workflow Specification (placeholder)'
-stepsCompleted: ['step-00-conversion', 'step-02-classification', 'step-03-requirements', 'step-04-tools', 'step-05-plan-review', 'step-06-design', 'step-07-foundation', 'step-08-build-step-01', 'step-09-build-all-steps']
-created: 2026-02-26
-status: VALIDATED
-approvedDate: 2026-02-26
-validationDate: 2026-02-26
-validationStatus: PASS
-validationReport: './validation-report.md'
----
-
-# Workflow Creation Plan
-
-## Conversion Source
-
-**Original Path:** src/workflows/test-skill/test-skill.spec.md
-**Original Format:** BMAD Workflow Specification (placeholder — not yet implemented)
-**Detected Structure:** Single specification document defining a 6-step create-only workflow
-
----
-
-## Original Workflow Analysis
-
-### Goal (from source)
-
-Cognitive completeness verification — Verifies that a skill is complete enough to be useful to an AI agent. Supports two modes: naive (individual skill API surface coverage) and contextual (stack skill coherence validation). Produces a completeness score and gap report. Acts as a quality gate before export.
-
-### Original Steps (Complete List)
-
-**Step 1:** load-skill - Load skill and metadata
-**Step 2:** detect-mode - Determine naive (individual) vs contextual (stack) testing mode
-**Step 3:** coverage-check - Compare documented exports against source API surface
-**Step 4:** coherence-check - Validate internal consistency (stack: references match SKILL.md)
-**Step 5:** score - Calculate completeness score
-**Step 6:** gap-report - Generate gap report with specific remediation suggestions
-
-### Output / Deliverable
-
-- Completeness score (pass/fail with percentage)
-- Gap report with specific missing items and remediation suggestions
-
-### Input Requirements
-
-**Required:**
-- Skill path (`{skills_output_folder}/{name}/`)
-
-**Optional:**
-- `--batch` flag with list of skills or directory
-- Source path override
-
-### Key Instructions to LLM
-
-- Spec assigns **Ferris (Audit mode)** as the primary agent
-- Create-only mode (steps-c/ only) — no edit or validate modal variants
-- Document-producing workflow
-- Must support `--batch` for testing multiple skills
-- Quality gate role: determines if a skill is ready for export
-- Two distinct analysis modes (naive vs contextual) require branching logic
-
----
-
-## Conversion Notes
-
-**What works well in original:**
-- Clear step decomposition with well-defined goals
-- Dual-mode architecture (naive/contextual) is well-scoped
-- Input/output requirements are explicit
-- Quality gate purpose is clear
-- Correct lifecycle positioning: test-skill sits between create-skill and export-skill (Setup → Analyze → Brief → Create → **Test** → Export)
-- Ferris Audit mode assignment is architecturally correct
-
-**What needs improvement:**
-- No step-level detail — only names and one-line goals exist
-- No menu handling, user interaction points, or continuation flow defined
-- No frontmatter configuration for steps
-- No data files, templates, or output format specifications
-- Batch mode mechanics are unspecified
-- No forge tier awareness — coverage depth must scale with Quick/Forge/Forge+/Deep tiers
-- No reference to forge-tier.yaml sidecar state for tier detection
-- Zero hallucination principle not encoded — scoring must trace to AST citations
-
-**Compliance gaps identified:**
-- Missing: workflow.md entry point with proper frontmatter
-- Missing: Step files with BMAD step-file architecture
-- Missing: Frontmatter variables for configurable paths
-- Missing: Menu handling and user interaction patterns
-- Missing: State tracking via stepsCompleted
-- Missing: Output document template/format specification
-- Missing: Step processing rules and execution protocols
-
-**Memory-informed design additions:**
-- Tier-aware coverage: Quick=file/structure check, Forge=AST API surface, Deep=cross-repo coherence
-- Batch mode: --batch with --continue for sidecar-checkpointed resume (matches create-skill pattern)
-- Batch produces per-skill reports with optional consolidated summary
-- Configurable pass/fail threshold (default overridable)
-- Must read forge-tier.yaml from forger-sidecar to determine available analysis depth
-
----
-
-## Classification Decisions
-
-**Workflow Name:** test-skill
-**Target Path:** {project-root}/_bmad/skf/workflows/test-skill
-
-**4 Key Decisions:**
-1. **Document Output:** true (completeness score + gap report)
-2. **Module Affiliation:** SKF (Skill Forge)
-3. **Session Type:** single-session
-4. **Lifecycle Support:** create-only
-
-**Structure Implications:**
-- Only `steps-c/` folder needed (no steps-e/ or steps-v/)
-- Standard init (no step-01b-continue.md needed)
-- Free-form output template for test report document
-- Access to SKF module variables (forge tier config, skills paths)
-- Batch mode is a loop wrapper, not a multi-session concern
-
----
-
-## Requirements
-
-**Flow Structure:**
-- Pattern: branching (by naive/contextual mode at step 02)
-- Phases: Loading → Analysis → Scoring → Reporting
-- Estimated steps: 6
-- Branch point: Step 02 detect-mode determines naive vs contextual path
-- Step 04 (coherence-check) is conditional — full execution in contextual mode, simplified/skipped in naive mode
-
-**User Interaction:**
-- Style: mostly autonomous (Ferris Audit mode — deterministic analysis)
-- Decision points: none mid-flow; mode detection is automatic based on skill type
-- Checkpoint frequency: entry (input validation) and exit (report delivery) only
-
-**Inputs Required:**
-- Required: skill path (`{skills_output_folder}/{name}/`), forge tier state (`forger-sidecar/forge-tier.yaml`)
-- Optional: `--batch` flag with skill list or directory, `--continue` for interrupted batch resume, source path override, custom pass/fail threshold
-- Prerequisites: setup-forge completed (tier detected), skill created via create-skill (SKILL.md exists)
-
-**Output Specifications:**
-- Type: document (structured test report)
-- Format: structured (defined sections for machine-readability and batch comparison)
-- Sections: Test Summary, Coverage Analysis, Coherence Analysis (contextual only), Completeness Score, Gap Report, Metadata
-- Frequency: per-skill report; batch produces per-skill reports + optional consolidated summary
-
-**Success Criteria:**
-- Completeness score is accurate and traceable to code via AST citations (zero hallucination)
-- Every gap has specific, actionable remediation suggestion
-- Pass/fail threshold respected and clearly reported
-- Naive mode correctly limits scope to API surface coverage
-- Contextual mode validates cross-reference coherence for stack skills
-- Batch mode produces consistent per-skill reports
-- Report format parseable for downstream automation (export-skill can read it)
-
-**Instruction Style:**
-- Overall: prescriptive (audit/testing — exact procedures per check)
-- Notes: Ferris Audit mode operates methodically; coverage check has specific comparison logic, coherence check has specific validation rules, scoring has defined formula
-
----
-
-## Tools Configuration
-
-**Core BMAD Tools:**
-- **Party Mode:** excluded — deterministic audit, no creative exploration
-- **Advanced Elicitation:** excluded — no facilitative questioning needed
-- **Brainstorming:** excluded — no idea generation in testing/scoring workflow
-
-**LLM Features:**
-- **Web-Browsing:** excluded — all analysis is local (source code + SKILL.md)
-- **File I/O:** included — read SKILL.md, metadata.json, forge-tier.yaml, source files; write test report
-- **Sub-Agents:** included — coverage-check and coherence-check as parallel sub-agents (batch efficiency)
-- **Sub-Processes:** included — batch mode parallel skill testing
-
-**Memory:**
-- Type: single-session (no continuation tracking)
-- Sidecar: read-only access to forge-tier.yaml for tier detection; batch checkpointing for --continue
-
-**External Integrations:**
-- **ast-grep:** required for Forge/Deep tier — AST-backed API surface comparison in coverage-check
-- **QMD:** optional for Deep tier — cross-reference knowledge search for coherence enrichment
-- **gh CLI:** optional for Deep tier — cross-repo reference verification
-
-**Installation Requirements:**
-- All external tools managed by forge tier system (setup-forge detects availability)
-- No additional installation — test-skill gracefully degrades based on detected tier
-
-**Structure Preview:**
-- Phase 1: Initialization — validate skill path, read forge tier, load metadata
-- Phase 2: Mode Detection — determine naive vs contextual based on skill type
-- Phase 3: Analysis — coverage-check (all tiers) + coherence-check (contextual only), depth scales with tier
-- Phase 4: Scoring & Reporting — calculate score, apply threshold, generate gap report, write test report
-
----
-
-## Workflow Design
-
-### Step Sequence
-
-| # | Name | Type | Menu | Goal |
-|---|------|------|------|------|
-| 01 | init | Init with Input Discovery | Auto-proceed | Discover skill, load forge tier, validate inputs |
-| 02 | detect-mode | Middle (Simple) | Auto-proceed | Determine naive vs contextual from skill metadata |
-| 03 | coverage-check | Validation Sequence | Auto-proceed | Compare documented exports against source API surface |
-| 04 | coherence-check | Validation Sequence (conditional) | Auto-proceed | Validate references/coherence (contextual=full, naive=simplified) |
-| 04b | external-validators | Middle (Simple) | Auto-proceed | Run external validation tools (skill-check, tessl) and capture scores |
-| 05 | score | Middle (Simple) | Auto-proceed | Calculate completeness score from findings |
-| 06 | report | Final Step | C only | Generate gap report, finalize document, recommend next workflow |
-
-### Interaction Flow
-
-```
-User provides skill path → [01] init (auto) → [02] detect-mode (auto) →
-[03] coverage-check (auto) → [04] coherence-check (auto) →
-[04b] external-validators (auto) → [05] score (auto) →
-[06] report → [C] finalize
-```
-
-### Data Flow
-
-- Step 01: loads SKILL.md, metadata.json, forge-tier.yaml → creates output from template
-- Step 02: reads metadata type → sets {testMode} naive|contextual → appends Test Summary
-- Step 03: reads SKILL.md exports + source files → subprocess AST analysis → appends Coverage Analysis
-- Step 04: reads SKILL.md references → conditional depth by mode → appends Coherence Analysis
-- Step 04b: runs skill-check and tessl against skill directory → captures external scores → appends External Validation
-- Step 05: reads coverage + coherence + external validation findings → calculates score → appends Completeness Score
-- Step 06: reads all findings → generates remediation per gap → appends Gap Report → finalizes
-
-### Subprocess Optimization
-
-| Step | Pattern | What | Returns |
-|------|---------|------|---------|
-| 03 | Pattern 2 (deep analysis) | Per-source-file AST export comparison | {file, exports_found, exports_documented, missing, extra} |
-| 03 | Pattern 4 (parallel) | Batch mode: parallel per-skill testing | Per-skill coverage findings |
-| 04 | Pattern 1 (grep/regex) | Grep skill files for references | {references_found, files_checked, broken_refs} |
-| 04 | Pattern 2 (deep analysis) | Per-reference coherence validation | {ref, target_exists, type_matches, signature_matches} |
-
-Fallback: All subprocesses have main-thread fallback.
-
-### File Structure
-
-```
-test-skill/
-├── workflow.md
-├── data/
-│   ├── scoring-rules.md
-│   └── output-section-formats.md
-├── steps-c/
-│   ├── step-01-init.md
-│   ├── step-02-detect-mode.md
-│   ├── step-03-coverage-check.md
-│   ├── step-04-coherence-check.md
-│   ├── step-04b-external-validators.md
-│   ├── step-05-score.md
-│   └── step-06-report.md
-└── templates/
-    └── test-report-template.md
-```
-
-### Role & Persona
-
-Ferris (Audit mode): methodical, precise, evidence-based. Zero hallucination — every finding cites AST evidence with file:line references. Tier-aware — states analysis depth in report. Non-judgmental — reports facts, suggests remediation.
-
-### Workflow Chaining
-
-- **Previous:** create-skill (required — SKILL.md must exist)
-- **Next:** export-skill (if pass) or update-skill (if fail)
-- **Input contract:** `{skills_output_folder}/{name}/SKILL.md` + `metadata.json` + `{sidecar_path}/forge-tier.yaml`
-- **Output contract:** `{forge_data_folder}/{skill_name}/test-report-{skill_name}.md` with frontmatter: workflowType, testResult, score, nextWorkflow
-
-### Output Template (Structured)
-
-```markdown
----
-workflowType: 'test-skill'
-skillName: '{skill_name}'
-testMode: '{naive|contextual}'
-forgeTier: '{Quick|Forge|Forge+|Deep}'
-testResult: '{pass|fail}'
-score: '{N%}'
-threshold: '{N%}'
-testDate: '{date}'
-nextWorkflow: '{export-skill|update-skill}'
----
-
-# Test Report: {skill_name}
-
-## Test Summary
-## Coverage Analysis
-## Coherence Analysis
-## Completeness Score
-## Gap Report
-```