docs-dry-refactorlisted

Operationalises spec/project/mkdocs-structure/ §Snippet inclusion (DRY). Detects paragraph duplication across MkDocs pages and, with per-snippet user approval, extracts duplicates into `mkdocs-include-markdown-plugin` includes pointing at a canonical source (preferring a live source file over a dedicated per-language _snippets/ folder). Three operations: `audit` (read-only ranked findings), `propose` (surface canonical source, markers, include directives for a target snippet ID, await approval), `apply` (write markers, replace consumer blocks, verify via `mkdocs build --strict`). Invoke when the user asks to dedupe, DRY-refactor, extract snippets, or factor out duplicated MkDocs content; also handles equivalent German-language requests. Don't use for non-MkDocs markdown trees, single-file snippet authoring, prose linting (`prose-vale-curator`), structural scaffolding (`mkdocs-structure-apply`), or drift detection (`docs-freshness-checker`). Supports resume on re-invocation per `spec/claude/resumable-work/`.
nolte/claude-shared · ★ 0 · Code & Development · score 76

Install: claude install-skill nolte/claude-shared

# Docs DRY Refactor Operationalises `spec/project/mkdocs-structure/<canonical_language>.md` §Snippet inclusion (DRY) inside the current repository. The skill scans the configured `docs/<lang>/` trees for paragraph-level duplication that violates the spec's DRY threshold, proposes a canonical-source extraction for each finding, and—after explicit per-snippet user approval—rewrites the consumer pages to include from that source via `mkdocs-include-markdown-plugin`. When the spec isn't present in the target repository, fall back to the copy shipped by the `nolte-shared` plugin (read it at runtime from `${CLAUDE_PLUGIN_ROOT}/spec/project/mkdocs-structure/<canonical_language>.md`). Never invent requirements that don't appear in the spec. ## Why this is a skill, not an agent Per `spec/claude/skill-vs-agent/` §Decision dimensions, this capability is a skill because: - **Mid-flow per-snippet user approval is the contract.** Each candidate extraction (canonical source choice, marker placement, every consumer's include directive) is a separate user decision; an agent's fire-and-forget shape would collapse the per-snippet approvals into a single opaque report. - **Persistent on-disk output that flows back into the main conversation.** The findings table, the per-finding proposal, and the build-verification output all surface in the conversation so the user can decide; isolating them in a structured-report boundary would obscure the per-snippet approval surface. - **Orchestrator pat