bmad-module-skill-forge 0.6.0 → 0.7.1

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

@@ -1,263 +0,0 @@
----
-validationDate: 2026-02-26
-workflowName: export-skill
-workflowPath: _bmad-output/bmb-creations/workflows/export-skill
-validationStatus: PASS
----
-
-# Validation Report: export-skill
-
-**Validation Started:** 2026-02-26
-**Validator:** BMAD Workflow Validation System
-**Standards Version:** BMAD Workflow Standards
-
----
-
-## File Structure & Size
-
-**Status: ✅ PASS**
-
-### Folder Structure
-
-```
-export-skill/
-├── workflow.md                          ✅ Entry point present
-├── data/                                ✅ Shared data folder
-│   ├── snippet-format.md               ✅ ADR-L template
-│   └── managed-section-format.md       ✅ ADR-J template
-├── steps-c/                             ✅ Create steps folder
-│   ├── step-01-load-skill.md           ✅ Sequential
-│   ├── step-02-package.md              ✅ Sequential
-│   ├── step-03-generate-snippet.md     ✅ Sequential
-│   ├── step-04-update-context.md       ✅ Sequential
-│   ├── step-05-token-report.md         ✅ Sequential
-│   └── step-06-summary.md             ✅ Sequential (final)
-├── workflow-plan-export-skill.md        ✅ Build plan
-└── validation-report.md                 ✅ This report
-```
-
-- ✅ workflow.md exists
-- ✅ Steps organized in steps-c/ folder
-- ✅ Data files organized in data/ folder
-- ✅ Sequential numbering (01-06), no gaps
-- ✅ Final step exists (step-06-summary.md)
-- ✅ No step-01b-continue.md (correct — single-session workflow)
-
-### File Size Analysis
-
-| File | Lines | Limit | Status |
-|------|-------|-------|--------|
-| step-01-load-skill.md | 185 | 200 | ✅ Good |
-| step-02-package.md | 153 | 200 | ✅ Good |
-| step-03-generate-snippet.md | 166 | 200 | ✅ Good |
-| step-04-update-context.md | 220 | 250 | ⚠️ Approaching limit |
-| step-05-token-report.md | 126 | 200 | ✅ Good |
-| step-06-summary.md | 167 | 200 | ✅ Good |
-| workflow.md | 58 | — | ✅ Good |
-| data/snippet-format.md | 32 | — | ✅ Good |
-| data/managed-section-format.md | 70 | — | ✅ Good |
-
-**Note:** step-04-update-context.md at 220 lines is approaching the 250-line max but within limits. This is justified by the three-case CLAUDE.md logic complexity.
-
-### Issues Found: 0
-
-## Frontmatter Validation
-
-**Status: ✅ PASS (after remediation)**
-
-### Findings
-
-3 unused frontmatter variable violations found and remediated:
-
-| File | Variable | Issue | Fix |
-|------|----------|-------|-----|
-| workflow.md | `web_bundle: true` | Not referenced in body | Removed |
-| workflow.md | `installed_path` | Not referenced in body | Removed |
-| step-01-load-skill.md | `snippetFormatData` | Not used in step-01 (belongs in step-03 only) | Removed |
-
-All remaining frontmatter variables verified as used:
-- `name`, `description` — present in all files ✅
-- `nextStepFile` — correctly referenced in steps 01-05, absent from step-06 (final) ✅
-- `snippetFormatData` — step-03 only ✅
-- `managedSectionData` — step-04 only ✅
-
-### Issues Found: 3 (all remediated)
-
----
-
-## Critical Path Violations
-
-**Status: ✅ PASS**
-
-- All step file references resolve to existing files
-- `nextStepFile` chain: step-01 → 02 → 03 → 04 → 05 → 06 (verified)
-- Data file references (`snippetFormatData`, `managedSectionData`) resolve correctly
-- No circular references or broken paths detected
-
-### Issues Found: 0
-
----
-
-## Menu Handling Validation
-
-**Status: ✅ PASS (with notes)**
-
-| Step | Pattern | Expected | Actual | Status |
-|------|---------|----------|--------|--------|
-| step-01 | P2: C only | Confirmation gate | C-only menu with halt | ✅ |
-| step-02 | P3: Auto-proceed | No user choice | Auto-proceed after validation | ✅ |
-| step-03 | P3: Auto-proceed | No user choice | Auto-proceed after snippet gen | ✅ |
-| step-04 | P2: C only | Confirmation gate | C-only with dry-run auto-proceed | ✅ |
-| step-05 | P3: Auto-proceed | No user choice | Auto-proceed after report | ✅ |
-| step-06 | Final | No next step | No menu, workflow ends | ✅ |
-
-**Notes:**
-- step-04 dry-run path auto-proceeds without a formal Menu Handling Logic block — acceptable design choice for conditional behavior
-- step-02 auto-proceed message says "snippet generation" but step-03 may be skipped (passive_context off) — technically accurate since step-03 is loaded next regardless and handles the skip internally
-
-### Issues Found: 0
-
----
-
-## Step Type Validation
-
-**Status: ✅ PASS (with notes)**
-
-| Step | Declared Type | Actual Behavior | Match |
-|------|--------------|-----------------|-------|
-| step-01 | Init (Input Discovery) | Discovers skill, validates, confirms | ✅ |
-| step-02 | Middle (Simple) | Validates package, auto-proceeds | ✅ |
-| step-03 | Middle (Simple) | Generates snippet, auto-proceeds | ✅ |
-| step-04 | Middle (Standard) | Three-case logic, confirmation gate | ✅ |
-| step-05 | Middle (Simple) | Token report, auto-proceeds | ✅ |
-| step-06 | Final | Summary, no next step | ✅ |
-
-**Notes:**
-- step-04 uses P2 C-only menu rather than P1 standard menu — matches design spec's "confirmation gate" pattern, appropriate for shared file modification
-
-### Issues Found: 0
-
----
-
-## Instruction Style Check
-
-**Status: ✅ PASS (after remediation)**
-
-**Overall Style:** Prescriptive (consistent across all steps)
-
-| File | Clarity Score | Notes |
-|------|--------------|-------|
-| workflow.md | 9/10 | Clear initialization, micro-file principles well articulated |
-| step-01-load-skill.md | 9/10 | Six numbered sections, clear conditionals, table summaries |
-| step-02-package.md | 9/10 | Package structure diagram, three readiness levels |
-| step-03-generate-snippet.md | 9/10 | Passive context check, snippet templates, edge cases handled |
-| step-04-update-context.md | 9/10 | Three-case logic, diff preview, post-write verification |
-| step-05-token-report.md | 8/10 | Consistent heuristic, clear budget impact breakdown |
-| step-06-summary.md | 9/10 | Conditional distribution instructions, clear completion signal |
-
-### Findings
-
-1. **workflow.md typo** — "adhere too 1 file" → fixed to "adhere to 1 file" ✅ Remediated
-2. **step-02 auto-proceed message** — says "snippet generation" without noting potential skip. Technically accurate (step-03 handles skip internally). Informational only.
-
-### Issues Found: 1 (remediated)
-
----
-
-## Collaborative Experience Check
-
-**Status: ✅ PASS**
-
-- **Role Clarity:** Clear — Ferris Delivery mode consistently defined across all steps
-- **User Interaction:** Good — 2 confirmation gates (step-01, step-04) at appropriate high-impact decision points
-- **Cross-Step Consistency:** Consistent — all `nextStepFile` references correct, data flow accurate, CONTEXT BOUNDARIES sections properly describe available data
-- **Dry-Run Handling:** Consistent across all relevant steps (03, 04 write-capable; 01, 02, 05, 06 read-only)
-- **Passive Context Handling:** Consistent — detected in step-01, gated in steps 03-04, reflected in step-05/06 output
-
-### Issues Found: 0
-
----
-
-## Subprocess Optimization Opportunities
-
-**Status: ✅ N/A**
-
-No subprocesses needed — lightweight linear pipeline with small files (~50-80 tokens per snippet). All operations are sequential with data dependencies between steps.
-
-### Issues Found: 0
-
----
-
-## Cohesive Review
-
-**Status: ✅ PASS (after remediation)**
-
-### End-to-End Flow: Coherent
-
-Complete logical progression: Load → Validate Package → Generate Snippet → Update Context → Report Tokens → Summary. No gaps between steps. Auto-proceed (02, 03, 05) and confirmation gates (01, 04) appropriately placed.
-
-### Data Consistency: Consistent
-
-- Variable names (`{skills_output_folder}`, `{skill-name}`, `{platform}`, `{target-file}`, `{forge_data_folder}`) used consistently across all steps
-- Data file content matches step usage: `snippet-format.md` ↔ step-03, `managed-section-format.md` ↔ step-04
-- ADR references: ADR-J (step-04), ADR-K (workflow.md), ADR-L (step-03) — all consistent
-
-### Edge Case Handling: Complete (after fixes)
-
-| Edge Case | Handling | Status |
-|-----------|----------|--------|
-| `--dry-run` | Writes → previews across steps 03, 04; reminder in step-06 | ✅ |
-| `passive_context: false` | Steps 03-04 skip; step-05 notes N/A; step-06 shows "skipped" | ✅ |
-| Missing files / invalid JSON | Hard halt in step-01 | ✅ |
-| Empty exports array | Graceful warn in step-01; omit exports line in step-03 | ✅ |
-| Zero skills during index rebuild | Generate header-only managed section (added) | ✅ Remediated |
-| Metadata validation gap (step-01 vs step-02) | Step-01 now validates all 7 required fields | ✅ Remediated |
-
-### Quality Score: 8.5/10
-
-### Top Strengths:
-1. Comprehensive passive_context and dry-run handling with consistent patterns across all steps
-2. Excellent data file design — self-contained templates for snippet format and managed section
-3. Strong safety model for shared file writes — diff preview, confirmation gate, post-write verification
-
-### Issues Found: 2 (both remediated)
-
----
-
-## Plan Quality Validation
-
-**Status: ✅ PASS**
-
-The workflow plan (`workflow-plan-export-skill.md`) accurately reflects the built workflow:
-- All 6 steps match plan design
-- Step types, menu patterns, and action plans align with implementation
-- Data flow matches step sequence
-- Error handling table covers all scenarios
-- File structure matches actual output
-
-### Issues Found: 0
-
----
-
-## Summary
-
-**Overall Status: ✅ PASS**
-
-| Check | Status | Issues | Remediated |
-|-------|--------|--------|------------|
-| File Structure & Size | ✅ PASS | 0 | — |
-| Frontmatter Validation | ✅ PASS | 3 | 3 ✅ |
-| Critical Path Violations | ✅ PASS | 0 | — |
-| Menu Handling | ✅ PASS | 0 | — |
-| Step Type Validation | ✅ PASS | 0 | — |
-| Instruction Style | ✅ PASS | 1 | 1 ✅ |
-| Collaborative Experience | ✅ PASS | 0 | — |
-| Subprocess Optimization | N/A | 0 | — |
-| Cohesive Review | ✅ PASS | 2 | 2 ✅ |
-| Plan Quality | ✅ PASS | 0 | — |
-
-**Total Issues Found:** 6
-**Total Remediated:** 6
-**Outstanding Issues:** 0
-
-**Validation Complete.** The export-skill workflow passes all validation checks. All discovered issues were remediated during the validation phase.

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

@@ -1,369 +0,0 @@
----
-conversionFrom: 'src/workflows/export-skill/export-skill.spec.md'
-originalFormat: 'Workflow Specification (pre-implementation blueprint)'
-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-steps-02-06', 'step-10-confirmation', 'step-11-completion']
-created: 2026-02-26
-completionDate: 2026-02-26
-status: COMPLETE
-approvedDate: 2026-02-26
-confirmationDate: 2026-02-26
-confirmationType: conversion
-coverageStatus: complete
-validationStatus: PASS
----
-
-# Workflow Creation Plan
-
-## Conversion Source
-
-**Original Path:** src/workflows/export-skill/export-skill.spec.md
-**Original Format:** Workflow Specification — pre-implementation blueprint (no step files exist yet)
-**Detected Structure:** 6-step create-only utility workflow defined as a spec with planned steps, inputs, outputs, and implementation notes. Enriched with module brief ADR decisions (J, K, L) and dual-output delivery strategy.
-
----
-
-## Original Workflow Analysis
-
-### Goal (from source)
-
-Package a skill for distribution + platform-aware context injection. Final delivery workflow that packages a skill as an agentskills.io-compliant package ready for `npx skills publish`. Generates/updates context-snippet.md. Updates the managed `<!-- SKF:BEGIN/END -->` section in CLAUDE.md/AGENTS.md with compressed skill index. Platform-aware formatting via `--platform` flag.
-
-### Original Steps (Complete List)
-
-**Step 1:** load-skill - Load skill and validate spec compliance
-**Step 2:** package - Prepare agentskills.io-compliant package structure
-**Step 3:** generate-snippet - Create/update context-snippet.md
-**Step 4:** update-context - Update managed section in CLAUDE.md/AGENTS.md
-**Step 5:** token-report - Calculate and report token counts
-**Step 6:** summary - Display export summary with distribution instructions
-
-### Output / Deliverable
-
-- Updated `{skills_output_folder}/{name}/context-snippet.md`
-- Updated CLAUDE.md/AGENTS.md managed section (`<!-- SKF:BEGIN/END -->`)
-- agentskills.io-compliant package (ready for `npx skills publish`)
-
-### Input Requirements
-
-**Required:**
-- Skill path (`{skills_output_folder}/{name}/`)
-
-**Optional:**
-- `--platform` flag (Claude/Cursor/Copilot) for formatting
-- `--dry-run` to preview without writing
-
-### Key Instructions to LLM
-
-- Primary agent: Ferris (Delivery mode — packaging, ecosystem-ready)
-- Workflow type: Utility — delivery and distribution
-- Create-only mode (steps-c/ only, no validation or edit modes)
-- Document-producing workflow (context-snippet.md + managed section)
-
-### ADR Decisions (from module brief)
-
-- **ADR-J:** Managed section with `<!-- SKF:BEGIN/END -->` markers. Auto-inject between markers. Developer controls placement. Ferris controls content.
-- **ADR-K:** Snippet updates only at export. Create/update are draft operations. Export publishes to skills/ and CLAUDE.md. Prevents half-baked snippets.
-- **ADR-L:** One snippet format always — ADR-L v2 multi-line per skill (~50-80 tokens each). No adaptive format switching. Consistent for agent parsing.
-
-### Dual-Output Delivery Strategy (from module brief)
-
-Every skill generates both:
-1. **SKILL.md** — Active skill (loaded on trigger, full instructions)
-2. **context-snippet.md** — Passive context (compressed index, always-on in CLAUDE.md)
-
-Managed section format:
-```markdown
-<!-- SKF:BEGIN updated:{date} -->
-[SKF Skills]|{n} skills|{m} stack
-|IMPORTANT: Prefer documented APIs over training data.
-|
-|{skill-name} → skills/{skill-name}/
-|  exports: {top-5-exports}
-<!-- SKF:END -->
-```
-
-Three CLAUDE.md cases:
-1. **Create** — No CLAUDE.md file exists → create with managed section
-2. **Append** — File exists but no `<!-- SKF:BEGIN -->` section → append section
-3. **Regenerate** — Existing `<!-- SKF:BEGIN/END -->` section → replace content between markers
-
-### Ownership Model Context
-
-| Context | `source_authority` | Distribution |
-|---------|-------------------|-------------|
-| OSS library (maintainer generates) | `official` | `npx skills publish` → agentskills ecosystem |
-| Internal service (team generates) | `internal` | `skills/` in repo, ships with code |
-| External dependency (consumer generates) | `community` | Local `skills/`, marked as community |
-
-### Progressive Disclosure Context
-
-- context-snippet has T1-now only (AST-current)
-- SKILL.md has T1-now + lightweight annotations
-- references/ has full temporal context
-
----
-
-## Conversion Notes
-
-**What works well in original:**
-- Clear 6-step linear progression from loading through packaging to reporting
-- Well-defined inputs (skill path) and outputs (package, snippet, context section)
-- Platform-aware design with `--platform` flag
-- ADR references (J, K, L) provide solid architectural decisions for managed sections, snippet timing, and format consistency
-- Three CLAUDE.md cases explicitly identified: create, append, regenerate
-- Position in workflow chain: downstream of create-skill/update-skill, final delivery step
-
-**What needs improvement:**
-- Spec is a blueprint only — no step files, templates, or data files exist yet
-- No detailed step-level instructions, menus, or execution protocols defined
-- No error handling or validation logic specified
-- No dry-run implementation details
-- No detail on agentskills.io package structure requirements
-- No specification of how platform-specific formatting differs between Claude/Cursor/Copilot
-
-**Compliance gaps identified:**
-- No step files exist (need full creation from spec)
-- No frontmatter with step metadata
-- No menu handling or continuation logic
-- No data files or templates defined (snippet format template, package structure template needed)
-- No subprocess/agent delegation patterns specified
-
----
-
-## Classification Decisions
-
-**Workflow Name:** export-skill
-**Target Path:** {project-root}/_bmad/skf/workflows/export-skill/
-
-**4 Key Decisions:**
-1. **Document Output:** false (action-based — produces files as side effects of packaging actions, not progressive document building)
-2. **Module Affiliation:** SKF (Skill Forge module — access to skills_output_folder, forge tier, sidecar state)
-3. **Session Type:** single-session (fast utility workflow — load, package, write, report)
-4. **Lifecycle Support:** create-only (steps-c/ only — deterministic utility, re-export to fix)
-
-**Structure Implications:**
-- `steps-c/` directory only (no steps-e/ or steps-v/)
-- Standard `step-01-init.md` (no continuation logic, no step-01b-continue.md)
-- No `stepsCompleted` tracking in output frontmatter
-- Action-based step pattern — steps perform file operations rather than building a document
-- Data folder needed for snippet format template and package structure reference
-- Ferris Delivery mode activated (only workflow using this mode)
-
----
-
-## Requirements
-
-**Flow Structure:**
-- Pattern: Linear (Step 1 → Step 2 → Step 3 → Step 4 → Step 5 → Step 6)
-- Phases: Load & Validate → Package & Generate → Publish → Report
-- Estimated steps: 6
-- No branching — three CLAUDE.md cases (create/append/regenerate) handled as conditional logic within step-04
-
-**User Interaction:**
-- Style: Mostly autonomous with targeted checkpoints
-- Decision points: After step-01 (confirm correct skill loaded), after step-04 (confirm CLAUDE.md changes)
-- Checkpoint frequency: 2 checkpoints in 6 steps
-- `--dry-run` flag converts entire workflow to preview mode (no writes)
-
-**Inputs Required:**
-- Required: Skill path (`{skills_output_folder}/{name}/`) containing SKILL.md, metadata.json
-- Optional: `--platform` flag (Claude/Cursor/Copilot), `--dry-run` flag
-- Prerequisites: Skill must exist (created by create-skill or update-skill). test-skill recommended but advisory.
-- Implicit: forge-config.yaml for `passive_context` opt-out check, existing CLAUDE.md/AGENTS.md content
-
-**Output Specifications:**
-- Type: Action-based (file writes + console output)
-- Files written:
-  1. `{skills_output_folder}/{name}/context-snippet.md` — ADR-L v2 format (~80-120 tokens/skill)
-  2. CLAUDE.md or AGENTS.md — managed `<!-- SKF:BEGIN/END -->` section per ADR-J (3 cases: create/append/regenerate)
-  3. agentskills.io package structure — ready for `npx skills publish`
-- Console output: Token count report + export summary with distribution instructions
-- Frequency: Single skill per run (batch mode deferred)
-
-**Success Criteria:**
-- Skill loaded and validated against agentskills.io spec
-- context-snippet.md generated with correct ADR-L v2 format
-- CLAUDE.md managed section correctly handles all three cases without corrupting existing content
-- Token counts calculated and reported for all generated artifacts
-- Package structure ready for `npx skills publish`
-- `--dry-run` shows preview with zero file writes
-- Zero data loss — existing CLAUDE.md content outside markers preserved
-
-**Instruction Style:**
-- Overall: Prescriptive (deterministic utility, exact file operations)
-- Notes: Auto-proceed between most steps (P3 pattern). Confirmation gates only at step-01 (skill loaded) and step-04 (CLAUDE.md changes). Similar to create-skill's compile/generate-artifacts prescriptive pattern.
-
----
-
-## Tools Configuration
-
-**Core BMAD Tools:**
-- **Party Mode:** Excluded — deterministic utility workflow, no creative exploration needed
-- **Advanced Elicitation:** Excluded — no complex decision-making requiring deep probing
-- **Brainstorming:** Excluded — no ideation phases
-
-**LLM Features:**
-- **File I/O:** Included — Core requirement. Reads skill artifacts, writes context-snippet.md, modifies CLAUDE.md/AGENTS.md, creates package files
-- **Web-Browsing:** Excluded — all data is local
-- **Sub-Agents:** Excluded — linear pipeline, no parallelization
-- **Sub-Processes:** Excluded — sequential dependency chain
-
-**Memory:**
-- Type: Single-session (no sidecar, no continuation tracking)
-- State: In-context only — skill metadata and generated content passed between steps
-
-**External Integrations:**
-- None required. All operations are local file system.
-
-**Installation Requirements:**
-- None. All tools are built-in LLM capabilities.
-
-## Workflow Structure Preview
-
-**Phase 1: Load & Validate (step-01-load-skill)**
-- Load skill from `{skills_output_folder}/{name}/`
-- Validate SKILL.md, metadata.json, references/ exist
-- Check agentskills.io spec compliance
-- Read metadata for skill type, source_authority, exports
-- **Checkpoint:** Confirm correct skill loaded, show metadata summary
-
-**Phase 2: Package (step-02-package)**
-- Build agentskills.io-compliant package structure
-- Validate package completeness
-- Auto-proceed
-
-**Phase 3: Generate Snippet (step-03-generate-snippet)**
-- Generate context-snippet.md in ADR-L v2 format
-- T1-now content only — top-5 exports, pointer to skill path
-- ~50-80 tokens per skill target
-- Auto-proceed
-
-**Phase 4: Update Context (step-04-update-context)**
-- Check `passive_context` opt-out in forge-config.yaml
-- Detect platform (Claude → CLAUDE.md, Cursor → .cursorrules, Copilot → AGENTS.md)
-- Execute three-case logic: create file / append section / regenerate section
-- Parse existing `<!-- SKF:BEGIN/END -->` markers, preserve all content outside
-- **Checkpoint:** Show diff of changes before writing
-
-**Phase 5: Token Report (step-05-token-report)**
-- Calculate token counts for context-snippet.md, managed section, SKILL.md, full package
-- Auto-proceed
-
-**Phase 6: Summary (step-06-summary)**
-- Display export summary: files written, token counts, distribution instructions
-- Show `npx skills publish` command if source_authority is `official`
-- Show internal distribution path if `internal`
-
----
-
-## Workflow Design
-
-### Step Sequence
-
-| Step | Name | Type | Menu Pattern | Goal |
-|------|------|------|-------------|------|
-| 01 | load-skill | Init (Input Discovery) | P2: C only | Load skill artifacts, validate spec compliance, confirm with user |
-| 02 | package | Middle (Simple) | P3: Auto-proceed | Build agentskills.io package structure |
-| 03 | generate-snippet | Middle (Simple) | P3: Auto-proceed | Generate context-snippet.md in ADR-L v2 format |
-| 04 | update-context | Middle (Standard) | P2: C only | Update managed section in CLAUDE.md/AGENTS.md (3-case logic) |
-| 05 | token-report | Middle (Simple) | P3: Auto-proceed | Calculate and report token counts |
-| 06 | summary | Final | No next step | Display export summary with distribution instructions |
-
-### Step Action Plans
-
-**Step 01 — load-skill (Init with Input Discovery):**
-- Discover skill at `{skills_output_folder}/{name}/`
-- Validate required files: SKILL.md, metadata.json
-- Validate optional: references/ directory
-- Read metadata.json for: skill_type, source_authority, exports list, generation date
-- Check forge-config.yaml for `passive_context` opt-out
-- Parse `--platform` flag (default: Claude → CLAUDE.md)
-- Parse `--dry-run` flag
-- Present skill summary to user, confirm correct skill
-- Menu: [C] Continue
-
-**Step 02 — package (Middle Simple, auto-proceed):**
-- Assemble agentskills.io package structure from skill artifacts
-- Validate package contains: SKILL.md, metadata.json, references/
-- Generate package manifest if needed
-- Verify metadata.json has required agentskills.io fields
-- Auto-proceed to step-03
-
-**Step 03 — generate-snippet (Middle Simple, auto-proceed):**
-- Load data/snippet-format.md template (ADR-L)
-- Generate context-snippet.md from metadata: skill name, path, top-5 exports
-- ADR-L v2 format, ~50-80 tokens target, T1-now content only
-- Write to `{skills_output_folder}/{name}/context-snippet.md` (or preview if dry-run)
-- Auto-proceed to step-04
-
-**Step 04 — update-context (Middle Standard, confirmation gate):**
-- Load data/managed-section-format.md template (ADR-J)
-- Determine target file based on platform flag (Claude→CLAUDE.md, Cursor→.cursorrules, Copilot→AGENTS.md)
-- Execute three-case detection: Create / Append / Regenerate
-- For regenerate: scan ALL skills in `{skills_output_folder}/*/context-snippet.md` to rebuild complete index
-- Show diff preview of changes
-- Menu: [C] Continue to write (or skip if dry-run)
-
-**Step 05 — token-report (Middle Simple, auto-proceed):**
-- Calculate token counts: context-snippet.md, managed section, SKILL.md, full package
-- Display token report table
-- Auto-proceed to step-06
-
-**Step 06 — summary (Final):**
-- Display export summary: files written/previewed, token counts
-- Distribution instructions based on source_authority (official/internal/community)
-- If dry-run: remind no files written, show command for real run
-- Recommend next: audit-skill, update-skill
-- No next step file
-
-### Data Flow
-
-```
-step-01 → skill_metadata, platform, dry_run, passive_context_enabled
-step-02 → package_structure (validated)
-step-03 → context-snippet.md written
-step-04 → CLAUDE.md/AGENTS.md updated
-step-05 → token_counts
-step-06 → console summary (terminal)
-```
-
-### File Structure
-
-```
-export-skill/
-├── workflow.md
-├── data/
-│   ├── snippet-format.md
-│   └── managed-section-format.md
-└── steps-c/
-    ├── step-01-load-skill.md
-    ├── step-02-package.md
-    ├── step-03-generate-snippet.md
-    ├── step-04-update-context.md
-    ├── step-05-token-report.md
-    └── step-06-summary.md
-```
-
-### Error Handling
-
-| Scenario | Response |
-|----------|----------|
-| Skill path doesn't exist | Hard halt — "Run create-skill first" |
-| SKILL.md or metadata.json missing | Hard halt — "Incomplete skill" |
-| passive_context: false | Skip steps 03-04, package-only export |
-| CLAUDE.md write failure | Hard halt, no partial writes |
-| --dry-run mode | All writes become preview-only |
-| No exports in metadata | Graceful — snippet shows name + path only |
-
-### Subprocess Optimization
-
-None needed — lightweight linear pipeline, small files (~50-80 tokens per snippet).
-
-### Workflow Chaining
-
-- **Prerequisites:** create-skill or update-skill (required), test-skill (advisory)
-- **Downstream:** None required — terminal workflow. Optional: audit-skill, update-skill
-
-### Role and Persona
-
-Ferris Delivery mode — precise, efficient, ecosystem-ready packaging. Brief forge metaphor at completion.