Capability: capability-docs
This layer itself: living capability documentation as the organized view of the per-work-item specs — dogfooded by documenting itself here.
What it is
The knowledge layer under workflow.capabilitiesDir (default docs/capabilities/) that answers "what does the product do today, by topic?" without making the reader merge specs in ticket order. Raw specs are deltas; capability docs are state.
Current behaviour
- Each capability SHALL have one living doc
<capabilitiesDir>/<capability>.md(template${CLAUDE_PLUGIN_ROOT}/skills/the-loop/templates/capability.md) containing: narrative, current behaviour (normative), design pointers, and a history table. - A capability doc SHALL be the single source of truth for the capability's current behaviour; the raw specs under
docs/specs/<id>/remain the historical record. - WHEN a work item changes a capability's behaviour THEN the loop SHALL update the affected capability doc(s) in the same PR, minting new docs for first-touched capabilities and updating the
capabilities.mdindex. - WHEN the ready-to-ship gate is evaluated THEN it SHALL require the fold-in (or "none affected" recorded in the execution log).
- Every behaviour statement SHALL be traceable via the history table to the work-item spec, decision record(s) and PR that produced it.
- The taxonomy SHALL admit both product-feature and architecture shaped capabilities and SHALL evolve through PR-review feedback on the docs themselves.
Design
docs/specs/issue-25/design.md · reference/workflow.md § capability docs
History
| Work item | What changed | Links |
|---|---|---|
| issue-25 | Introduced the layer: template, workflow.capabilitiesDir, fold-in gate, backfill of all existing capabilities | spec, decision-020, PR #26 |