Requirements Migrate

Convert existing monolithic requirements documents into the graph structure.

Process

1. Parse Source Document

Read the monolithic requirements file and extract:

• User stories (As a/I want/So that blocks)
• Acceptance criteria (Given/When/Then blocks)
• Functional requirements (REQ-XXX entries)
• Non-functional requirements (REQ-PERF/SEC/SCALE/UX entries)
• Scope (in/out/future)
• Risks, assumptions, open questions
• Dependencies

2. Plan Graph Structure

Map extracted content to graph nodes:

• Group stories by feature area (infer from headings or REQ prefixes)
• Each story becomes a US-NNN directory
• Each AC becomes an AC-NNN-{slug}.md file
• Functional requirements that duplicate stories are merged (not duplicated)
• NFRs become area-level NFR-NNN-{slug}.md files
• Decisions/rationale become _decisions/DEC-NNN-{slug}.md

3. Create Graph

For each area:
  mkdir -p requirements/{area}
  Write requirements/{area}/meta.md

  For each story in area:
    mkdir -p requirements/{area}/{US-NNN}
    Write requirements/{area}/{US-NNN}/meta.md

    For each AC in story:
      Write requirements/{area}/{US-NNN}/{AC-NNN}-{slug}.md

  For each NFR in area:
    Write requirements/{area}/{NFR-NNN}-{slug}.md

Write requirements/meta.md (root)
Write requirements/_scope.md
Write requirements/_risks.md
Write requirements/_questions.md

4. Verify

After migration:

• Run navigate lint to check structure
• Compare AC count: source doc vs graph nodes (should match)
• Check no content was lost (stories, criteria, NFRs)
• Verify frontmatter IDs match paths

5. Archive Source

Move the original monolith to requirements/_archived/ so it's
preserved but not the active source of truth.

Handling Ambiguity

• Stories without clear ACs: Create the story directory + meta.md,
  add a note in _questions.md that ACs need to be defined
• Functional reqs that aren't stories: If REQ-XXX doesn't map to a
  user story, create it as an AC under a synthetic story
• Inline NFRs: Extract performance/security mentions from story text
  into area-level NFR nodes
• Missing personas: Default to the most common persona in the doc,
  flag in _questions.md

Progressive Disclosure

• SKILL.md (this file): Migration process