← ClaudeAtlas

doc-enforcelisted

SDD documentation enforcement orchestrator. Runs the 15-row execution manifest against documentation/. Detects forbidden content, per-element budget violations (per-file caps deprecated in v2.0), within-section semantic issues, authoring-quality prose (weasel, unverifiable, missing-why), REQ-backlink gaps, doc source-anchor truth (Pass 15 — always runs). Conditionally invokes doc-enforce-lanes (per file in diff), doc-enforce-shape (api-reference / canonical lane files), and doc-enforce-truth (Implemented REQ docs or scope=all). Invoked by doc-updater on every PR-boundary trigger and by /sdd clean.
nikolanovoselec/codeflare · ★ 24 · Code & Development · score 69
Install: claude install-skill nikolanovoselec/codeflare
# Documentation Enforcement (orchestrator) This skill is the spine for SDD documentation enforcement. It runs the 15-row execution manifest against `documentation/` and orchestrates the conditional detail skills (`doc-enforce-lanes`, `doc-enforce-shape`, `doc-enforce-truth`). ## Inputs - `diff`: git diff against base (PR-boundary triggers) OR full-tree view (scope=all) - `scope`: `all` | `diff` (default `diff`) - `mode`: `interactive` | `auto` | `unleashed` (read from `sdd/spec/config.yml` when nested layout exists, else `sdd/config.yml` on flat layout) - `layout`: `nested` | `flat` (auto-detected via `test -d documentation/lanes`) **Layout-awareness.** All file globs in this skill respect the detected layout: - Lane files: `documentation/lanes/**/*.md` (nested) OR `documentation/*.md` excluding `README.md` (flat) - ADR ledger: `documentation/decisions/README.md` (both layouts; unchanged) - Triage / escalation: `documentation/.doc-coverage.md` (audit accumulator, still used to record Pass 12 cold-read gaps and Pass 15 retrofit-failure entries) ## Execution contract (binding) Every row of the manifest below MUST execute on every run. No cherry-picking; cost is never a valid skip. Manifest written FIRST with all rows `pending`, updated as each pass completes, finalised at run end. Pending rows at finalize emit HIGH `manifest-pending-at-finalize`. Status rows without concrete evidence counts emit HIGH `manifest-bare-evidence-count`. "skipped (looked clean)" is dishonest.