Requirements Navigate

Read, query, and maintain requirements graph structures. Generates and
refreshes meta.md files. Checks coverage and finds gaps.

Commands

Navigate (read graph)

Read a requirements node or area with progressive depth:

Depth 0 — frontmatter only (~5 tokens per node):

# Read just the frontmatter of all ACs in a story
for f in requirements/auth/US-001/AC-*.md; do
  head -20 "$f"  # frontmatter block
done

Depth 1 — meta.md (summary + tables):

# Read the story summary
cat requirements/auth/US-001/meta.md

Depth 2 — full content of specific nodes:

# Read a specific AC
cat requirements/auth/US-001/AC-001-login-success.md

Always start at depth 1 (meta.md) unless you need specific AC content.

Refresh meta.md

Regenerate a meta.md from its children when stale:

1. Read all sibling/child files in the directory
2. Parse frontmatter from each
3. Rebuild the summary table (counts, priorities, coverage stats)
4. Write updated meta.md preserving the id and type fields

Staleness check: Compare generated_at in meta.md frontmatter against
stat() mtimes of children. If any child is newer, regenerate.

When to refresh:

• Before displaying requirements status to user
• Before crew gate evaluation
• After batch AC creation
• On explicit user request

Coverage Query

Check which ACs have trace links:

For each AC in requirements/**/*.md where type: acceptance-criterion:
  - Has IMPLEMENTED_BY? → code coverage
  - Has TESTED_BY? → test coverage
  - Has VERIFIES? → verification coverage

Output format:

Coverage: auth
  US-001: 3/3 ACs, 1/3 implemented, 0/3 tested
  US-002: 2/2 ACs, 0/2 implemented, 0/2 tested
  Total: 5 ACs, 1 implemented (20%), 0 tested (0%)

Gap Analysis

Find structural problems:

• Orphan ACs: AC files not referenced in parent story meta.md
• Empty stories: Story directories with meta.md but no AC files
• Missing traces: P0 ACs without IMPLEMENTED_BY or TESTED_BY
• Broken traces: Trace targets that don't exist on disk
• Stale meta.md: Generated timestamps older than children
• ID mismatches: id: field doesn't match file path

Lint

Validate graph structure:

For each .md in requirements/**/:
  1. Has valid YAML frontmatter with required fields
  2. id: matches relative path from requirements/
  3. type: is a valid node type
  4. traces: targets exist (warn, don't fail — code may not exist yet)
  5. ACs have: priority, category, story
  6. Stories have: priority, complexity, persona, status
  7. meta.md tables match actual children

Fix mode: --fix auto-corrects:

• Regenerate stale meta.md files
• Update id: to match current path
• Add missing AC references to parent meta.md

Integration with traceability.py

The navigate skill reads graph nodes directly. For cross-phase queries
that span beyond requirements (design, code, tests, evidence),
traceability.py provides BFS traversal.

Bridge: When navigate finds IMPLEMENTED_BY or TESTED_BY traces
in frontmatter, these are the authoritative links for the requirement
layer. traceability.py can read these as an additional data source
alongside its DomainStore links.

Integration with search

Requirements graph nodes are indexable by search:index:

wicked-garden-search index requirements/

Enables queries like:

• "All P0 ACs without TESTED_BY" (frontmatter + FTS5)
• "All ACs tagged 'authentication'" (tag search)
• "Which stories trace to src/auth/" (reverse trace)

Progressive Disclosure

• SKILL.md (this file): Commands, query patterns, integration
• refs/meta-templates.md: Copy-paste meta.md templates for each level