Tasks: specs organization and capability documentation

Phase 3 of 3. Derived from the approved design.md. Docs/config-only work item: the "test" for each task is the repo's own quality gates (markdownlint, schema validation, pre-commit parity) — no production code, so no red→green cycle.

Task list

• [x] 1. Add .the-loop/templates/capability.md defining the capability-doc structure
  • Header, what-it-is, current behaviour (normative), design pointers, history table
  • Depends on: none
  • Requirements: R2.3, R6.1
  • Test: npx markdownlint-cli2 ".the-loop/templates/capability.md"
• [x] 2. Add workflow.capabilitiesDir to the config contract
  • config.schema.json (new key, default docs/capabilities), config.yaml, templates/config.yaml
  • Depends on: none
  • Requirements: R6.2
  • Test: uv run python scripts/validate_config.py
• [x] 3. Track the layer in .the-loop/manifest.yaml
  • Template entry + knowledge entries (capabilities.md index, <capability>.md records)
  • Depends on: 1
  • Requirements: R6.3
  • Test: manifest lists both; markdownlint/pre-commit green
• [x] 4. Wire the rule and gate into the operating model
  • skills/the-loop/SKILL.md (principle + knowledge section), skills/the-loop/reference/workflow.md (fold-in step + ready-to-ship gate item), README.md (rules + repository layout)
  • Depends on: 1, 2
  • Requirements: R4, R5, R6.4
  • Test: npx markdownlint-cli2 over changed files
• [x] 5. Backfill docs/capabilities/ from the existing specs
  • capabilities.md index + capability docs covering issue-1, issue-11, issue-12, issue-15, issue-17, issue-18, issue-21 and issue-25 (dogfood: this capability documents itself), each with history rows linking the specs/decisions
  • Depends on: 1, 4
  • Requirements: R1, R2, R3, R4, R7
  • Test: npx markdownlint-cli2 "docs/capabilities/*.md"
• [x] 6. Record docs/decisions/decision-020.md + index row
  • Capability docs as the organized view of specs and single source of truth for current behaviour
  • Depends on: 5
  • Requirements: R2 (decision trail)
  • Test: markdownlint green; decisions index links resolve
• [x] 7. Close out: execution log, phase label, evidence
  • Tick tasks here, update execution-log.md (phases, evidence), keep issue #25 label in sync, run the full gate suite
  • Depends on: 1–6
  • Requirements: all (evidence)
  • Test: pre-commit run --all-files (exactly what CI runs)

Dependency graph (DAG)

1 ─┬─→ 3
   ├─→ 4 ─→ 5 ─→ 6 ─→ 7
2 ─┘         ↑
        (1 also → 5)

Checkpoints

• After task 2: uv run python scripts/validate_config.py.
• After task 5: markdownlint over docs/capabilities/.
• After task 7 (final): pre-commit run --all-files — recorded in the execution log as the validation evidence.