Skills¶

Skills are reusable workflows Claude Code invokes via the Skill tool. In the claude-shared source tree they live at skills/<name>/SKILL.md; at runtime in a consuming project under .claude/skills/<name>/ or ~/.claude/skills/<name>/: or, as here, bundled inside the nolte-shared plugin.

Bundled skills¶

Skill	Description
`spec`	Manage multilingual specs (DE/EN), detect drift, maintain the index
`skill-management`	Scaffold or revise a skill under `skills/<name>/`
`skill-review`	Audit an existing skill against the authoring spec; persist findings as a review plan
`agent-review`	Audit an existing agent against the authoring spec; persist findings as a review plan
`pull-request-create`	Open a PR that conforms to the `pull-request-workflow` spec
`pull-request-merge`	Promote an open draft PR to merged on `develop`, applying labels and workflow gates
`release-notes-curate`	Augment the open release-drafter draft with a project-context section (per `release-skill-layer`)
`release-publish-trigger`	Validate pre-publish gates and dispatch `release-publish.yml` for the open draft on `develop`
`quality-gate`	Run lint + typecheck + tests in parallel and tabulate the results
`dependency-audit`	Scan dependencies for known CVEs (optionally licenses)
`project-structure-apply`	Audit the repo against the `project-structure` spec and scaffold missing artefacts
`github-issue-templates-apply`	Scaffold or update `.github/ISSUE_TEMPLATE/` per the `github-issue-templates` spec
`skill-agent-catalog-apply`	Wire up the MkDocs catalog that lists every skill and agent in a plugin repo
`mermaid-diagrams-apply`	Audit the MkDocs Mermaid setup against the `mermaid-diagrams` spec and emit spec-compliant diagrams
`vocab-drift-audit`	Diff repository-local Vale vocabulary against the pinned `nolte/vale-style` release
`workflow-health-triage`	Triage a failing GitHub Actions workflow per `workflow-health` spec; classify, dispatch a specialised agent, route the fix through the standard PR flow
`permission-allowlist-maintain`	Curate the committed `.claude/settings.json` `permissions.allow` list per `permission-allowlist` spec; user-gated per-entry approval
`audience-identify`	Identify the audiences of a bounded context and write a reviewable audience artifact
`mission-define`	First-write `project/mission.md` walking SMART one letter at a time
`mission-revise`	Edit `project/mission.md` and flip `mvp_status` lifecycle per the `mission` spec
`roadmap-init`	Scaffold initial `project/goals.md` and `project/roadmap.md` (Vision plus Outcomes plus empty queue)
`roadmap-refine`	Enforce the detail-level invariant in `project/roadmap.md` with structured violation reports
`roadmap-planner`	Add, promote, re-target roadmap items; flip the `mvp` flag
`feature-decompose`	Decompose a roadmap item into `project/features/<slug>.md` with `feature-consistency-reviewer` dispatch
`sprint-plan`	Create `project/sprints/<NNNN>-<slug>.md` with `value_statement` and `verifies_sprint_value`
`sprint-execute`	Drive the sprint lifecycle: `planned → active`, feature transitions, `last_commit` updates
`sprint-review`	Close a sprint with artefact validation, optionally chain into the release skill layer

Detail pages currently exist for skill-management and spec; more are being added. The authoritative source per skill is skills/<name>/SKILL.md in the source tree.

Skill anatomy¶

skills/<name>/
├── SKILL.md              # YAML frontmatter + instructions
├── templates/            # optional
├── references/           # optional
└── examples/             # optional

The frontmatter of every skill contains at minimum:

---
name: <folder-name>
description: Concrete trigger phrases, not abstract capabilities.
---

Full rules and acceptance criteria: Skill Authoring.