Assembly Checklist: {Product Name}

Assembly Checklist: {Product Name}

Order: {order-id} Client: {client-name} Product type: {product-type} Assembled by: Pablo (Production Line Architect) Date started: {start-date} Date completed: {completion-date}

Phase A: Repository Setup

  • Client repository created at {repo-path}
  • Directory structure matches work order specification
  • .claude/commands/ directory created
  • .claude/commands/templates/ directory created
  • content-in/ subdirectories created per domain requirements
  • Output directory structure created: {output-directory}/
  • .gitignore configured (excludes dev-only artifacts, sensitive files)
  • .claude/settings.json configured with shipping permissions
  • .claude/settings.local.json created for dev-only settings (will not ship)
  • Git initialized and initial commit made
  • Request folder template created (if applicable): {request-folder-pattern}

Directories created: __ / __ planned Date completed: {date} Notes: {notes}

Phase B: Core Agent Configuration

  • CLAUDE.md created with all 9 required sections
  • Section 1: Role definition -- product type, client name, capabilities, exclusions, language
  • Section 2: Workspace rules -- file safety, conventions, language rules, safety rules
  • Section 3: Structure -- directory layout matches actual repo
  • Section 4: Skills table -- all skills listed with command, input, output, model
  • Section 5: Workflow -- pipeline documented (or "independent skills" noted)
  • Section 6: Conventions -- file naming, output naming, request numbering, decision records
  • Section 7: Model selection -- cost optimization matrix configured per work order
  • Section 8: Permissions -- autonomous/confirm/deny table
  • Section 9: Known issues -- caveats and workarounds documented
  • README.md created (client-facing overview)
  • Working language correct throughout: {language}

Date completed: {date} Notes: {notes}

Phase C: Skill Assembly

  • All skills from solution spec built and placed in .claude/commands/
  • Each skill contains: frontmatter, input spec, processing steps, output spec, quality checks, model recommendation
  • Naming conventions match CLAUDE.md Section 6
  • Path resolution rules consistent across all skills
  • Model recommendations set per work order cost configuration

Skill Inventory

# Skill Name File Built Tested
1 {Skill name} {file-name}.md [ ] [ ]
2 {Skill name} {file-name}.md [ ] [ ]
3 {Skill name} {file-name}.md [ ] [ ]
{n} {Skill name} {file-name}.md [ ] [ ]

Skills built: __ / __ planned Skills tested (smoke test): __ / __ built Unplanned skills added: __ (document in notes) Date completed: {date} Notes: {notes}

Phase D: Template Library

  • All output templates from work order built and placed in .claude/commands/templates/
  • Templates use {placeholder} syntax for dynamic values
  • Templates match client visual and document standards
  • Templates are valid for their format (Markdown, Draw.io XML, etc.)

D.kit — AI-Layer Kit v1.0 integration (from WO-0010)

  • production-lines/digital-talent/templates/ai-layer-kit/ copied to {repo}/.claude/ai-layer/ (reference copy, not edited live)
  • CLAUDE-root.md placed at repo root as CLAUDE.md, placeholders filled
  • Per-domain CLAUDE-{domain}.md files created in .claude/ from CLAUDE-domain-template.md
  • hooks/stop-self-improve.sh installed in .claude/hooks/, executable, referenced in .claude/settings.json
  • Every skill in .claude/commands/*.md has a paths: frontmatter line (or documented justification for global load)
  • agents/explorer.md installed in .claude/agents/
  • Codebase-search MCP wired in .mcp.json (if applicable — sync with RD-0031 first)

Kit integration verified: [ ] Components installed: __ / 5 (5 = full v1)

Template Inventory

# Template Name File Built
1 {Template name} {file-name} [ ]
2 {Template name} {file-name} [ ]
{n} {Template name} {file-name} [ ]

Templates built: __ / __ planned Date completed: {date} Notes: {notes}

Phase E: Reference Materials

  • Domain framework documentation loaded into content-in/{domain}/
  • Tool documentation loaded into content-in/{tools}/ (if applicable)
  • Quality criteria and checklists loaded
  • All reference files in Markdown format (or Markdown summary alongside binary files)
  • File sizes within recommended limits

Reference Material Inventory

# Material Location Loaded
1 {Material name} content-in/{path} [ ]
2 {Material name} content-in/{path} [ ]
{n} {Material name} content-in/{path} [ ]

Reference files loaded: __ / __ planned Date completed: {date} Notes: {notes}

Phase F: Orchestration and Workflow

{If the product does not use orchestration, mark this phase as "N/A -- product uses independent skills" and skip to Phase G.}

  • Orchestrator skill built: .claude/commands/{orchestrator-command}.md
  • State detection logic implemented (scans for expected output files)
  • Step ordering matches CLAUDE.md Section 5 workflow
  • Mandatory halt points (human approval checkpoints) enforced
  • Progress tracking implemented
  • Workflow documentation written

Orchestration required: Yes / No Date completed: {date} Notes: {notes}

Phase G: Documentation

  • User Guide written -- covers all skills, workflow, troubleshooting
  • Configuration Guide written -- covers references, conventions, skills, templates, cost management
  • Skill Reference Card written -- quick-reference table for all skills
  • Quick-Start Cheat Sheet written -- one-page getting started guide
  • All documentation in correct language: {language}
  • All documentation references match actual file paths and skill names
  • User Guide contains the section « Qu'est-ce qui tourne dans votre talent ? » explaining the 5 AI-Layer Kit components (copy from ai-layer-kit/USAGE.md Section B)

Documents written: __ / 4 planned Date completed: {date} Notes: {notes}

Phase H: Verification

H1. Structural Verification

  • CLAUDE.md present with all 9 sections
  • README.md present
  • All skills from solution spec present in .claude/commands/
  • All templates present in .claude/commands/templates/
  • Reference materials present in content-in/
  • Orchestrator present (if applicable)
  • Client documentation complete (4 documents)

Structural result: PASS / FAIL Missing items: {list any missing items}

H2. Skill Smoke Tests

  • Each skill tested with sample input
  • Output format and naming verified for each skill
  • Domain terms and language verified in outputs
  • Error handling tested (missing input, invalid input)

Skills smoke-tested: __ / __ total Smoke test result: PASS / FAIL Failed skills: {list any failures}

H3. End-to-End Test (if orchestrator exists)

  • Test request created with sample input
  • Orchestrator executed through full pipeline
  • Each step produced expected output
  • Halt points triggered correctly
  • Resume after approval worked correctly
  • Final outputs complete and correct

E2E test result: PASS / FAIL / N/A Issues found: {list any issues}

H4. CLAUDE.md Consistency Check

  • Skills table (Section 4) matches actual .claude/commands/ files
  • File paths in Structure (Section 3) match actual directory layout
  • Naming conventions (Section 6) consistent with actual file names
  • Model recommendations (Section 7) match skill frontmatter

Consistency result: PASS / FAIL Discrepancies: {list any discrepancies}

H5. Dev Artifact Cleanup

  • Development-only files removed or excluded from deliverable
  • .claude/settings.local.json excluded (not shipped to client)
  • Internal notes, TODOs, and architecture analysis removed
  • No test data or sample outputs remaining in output directories

Cleanup result: PASS / FAIL

Overall verification date: {date} Notes: {notes}

Final Sign-Off

Field Value
Product {product-name}
Order {order-id}
Client {client-name}
Total skills {count}
Total templates {count}
Total reference files {count}
Total documentation pages {count}
All phases complete Yes / No
All verifications passed Yes / No
Ready for QA handoff Yes / No

Assembled by: Pablo (Production Line Architect) Sign-off date: {date} Signature: ___________________________