nolte
UserShared Claude Code skills and agents for consistent development workflows across projects.
Categories
Indexed Skills (50)
api-error-check
Statically checks a web API's error-handling surface for conformance against the project's own declared error contract — uniform error-body shape, populated required fields, a dynamically-generated error/correlation id, correct HTTP status-code semantics, and no internal-detail leakage (stack traces, raw driver messages, rendered queries, secrets). Detects the web framework (FastAPI / Flask / Django REST / Express / NestJS / Spring and comparable) and the error contract from project signals, falling back to RFC 9457 defaults only when none is declared. Invoke after adding or changing endpoints, before a release, or as a pre-PR error-handling gate; also on requests like "check the API error handling", "audit error responses", or equivalent German-language requests. Read-only: reports and recommends, never edits handler code. Don't use for the whole-codebase security audit (code-security-reviewer) or general code review (review skill).
dependency-audit
Scans the current project's dependency tree for known vulnerabilities (CVEs) and, when requested, license-compliance issues. Dispatches dependency-audit-scanner agent for the read-only scan step. Detects project kind from `pyproject.toml` / `requirements*.txt` / `poetry.lock` / `uv.lock` for Python and `package.json` / `package-lock.json` / `pnpm-lock.yaml` / `yarn.lock` for Node, runs the appropriate auditors, and produces a severity-sorted report with direct vs transitive attribution. Invoke when the user asks to "audit dependencies," "run a CVE scan," "check for vulnerable packages," "check the license allowlist," "run pip-audit," "run npm audit," or equivalent German-language requests. Also handles a pre-PR / pre-release dependency gate. Don't use for upgrading dependencies (an author's decision), for writing Renovate configs (`project-structure-apply`), or for the full license inventory / SPDX compliance process (`license-check`). Supports resume on re-invocation per `spec/claude/resumable-work/`.
deployment-chart-manage
Provisions and maintains a Helm application chart on the bjw-s `common` library, applying spec/project/bjw-s-common-chart-deployment/ and spec/project/kubernetes-deployment-best-practices/. Two operations: `provision` generates a complete best-practice chart greenfield; `reconcile` is the change-noticing process that detects what an application change (for example from `fullstack-developer`) requires the chart to gain, presents the gaps, and on approval extends the chart and re-audits. Dispatches `deployment-change-analyzer`, `bjw-common-deployment-generator`, and `deployment-bestpractices-reviewer`, and owns the chart files plus the README manual-steps section. Invoke to provision a deployment chart or reconcile one after the app changed; also German requests. Don't use to write application code (`fullstack-developer`) or run the CI quality gate (`quality-gate`). Supports resume per `spec/claude/resumable-work/`.
dockerfile-audit
Audits a project's Dockerfiles against spec/project/dockerfile-best-practices/ and produces a severity-classified audit artifact. The default `audit` operation dispatches the read-only dockerfile-audit-scanner agent, then hard-fails on any missing mandatory OCI core label (source, title, description, version, revision, created) present in neither the Dockerfile nor CI injection, and on any of the four mandatory non-label pillars (non-root numeric USER, in-layer secrets, base not pinned by tag+digest, missing .dockerignore); advisory pillars are scored. The opt-in `apply` operation inserts or merges the OCI LABEL block into the final stage. Invoke when the user asks to "audit a Dockerfile," "check OCI image labels," "run a Dockerfile best-practices check," or equivalent German-language requests. Don't use for dependency CVEs (dependency-audit), license inventory (license-check), or Kubernetes runtime hardening (deployment-bestpractices-reviewer). Supports resume per spec/claude/resumable-work/.
kpi-derive
Derives a business application's project-specific KPIs from its goals, requirement documents, and source code, writing a human-readable artifact to project/kpis/<slug>.md per spec/project/kpi-definition-process/. The `derive` operation dispatches the read-only kpi-signal-scanner agent, walks a GQM refinement (goal to question to metric), selects the key metrics as KPIs with the operator, and defines each against the SMART gate with a leading/lagging class; `revisit` re-derives against changed goals. Consumes an existing project/requirements/<slug>.md under a soft gate (warns and recommends requirements-elicit when absent, never blocks). Determination and definition only — never measurement, instrumentation, telemetry, or dashboarding. Invoke to derive KPIs, determine an app's KPIs, or define its key metrics; also German requests. Don't use to elicit requirements (requirements-elicit) or to build instrumentation or dashboards (fullstack-developer). Supports resume per spec/claude/resumable-work/.
license-check
Runs an end-to-end license-compliance check on the current project per spec/project/license-check/ and produce a license-check audit artifact. Dispatches license-check-scanner agent for the read-only inventory (SBOM with resolved licenses, SPDX identification, category classification), then applies the permissive-leaning allow/review/deny policy gate against the project's own outbound license, drives per-finding remediation (replace / exception with rationale / satisfy the obligation), verifies attribution/NOTICE and REUSE, records AI provenance, and writes the artifact under .audits/license-check/. Invoke when the user asks to "run a license check," "check license compliance," "audit licenses," "do a Lizenzcheck," "prüfe die Lizenzen," or for a pre-PR / pre-release license gate. Don't use for CVE / vulnerability scanning (that's dependency-audit) or for choosing the project's own outbound license. Supports resume on re-invocation per spec/claude/resumable-work/.
observability-audit
Audits an app's observability against spec/project/monitoring-observability/ and produces a severity-classified audit artifact plus an implementation plan for the gaps. The default `audit` operation dispatches the read-only observability-audit-scanner agent, then hard-fails on any missing-or-unwired mandatory pillar (metrics, structured logs, distributed traces, health + SLO + alert), the frontend and third-party floors, and the cardinality and PII-redaction guardrails; runtime-only behaviour is documented verify-at-runtime. The `plan` operation dispatches implementation-plan-author with the findings so a specialist (fullstack-developer) implements the instrumentation — there is no mechanical apply. Invoke to audit observability, check OTel instrumentation, or run an observability best-practices gate; also German. Don't use for Kubernetes probe wiring (deployment-bestpractices-reviewer), the PII/GDPR verdict (gdpr-data-protection-reviewer), or Dockerfile labels (dockerfile-audit). Supports resume.
release-regression-scope
Determines the release-relevant regression/E2E test scope from a release change-set per spec/project/release-regression-scope/, so a team runs targeted-but-safe regression before rollout. The default `scope` operation resolves the release range, dispatches the read-only release-regression-scope-scanner to attribute each change to its impacted areas via traceability inversion (change → requirement/TC-ID → verifying tests), selects the minimal tier/test set (E2E emphasised) covering those areas, widens non-attributable changes to worst-case full-area regression, and reports an auditable scope (in-scope areas, selected tests, exclusions + rationale, residual risk). A missing verifying test is a coverage-gap blocker. Invoke to scope release regression, pick targeted E2E, or gate a rollout on a fast subset; also German. Don't use to derive test cases (test-case-extractor), run tests (quality-gate), audit pyramid shape (test-pyramid-check), or drive the release (release-publish-trigger). Supports resume.
test-cycle-orchestrate
Drives one or more turns of the iterative test cycle defined in spec/project/test-cycle-foundation/ — case determination → execution → result analysis → code adaptation → re-execute — dispatching each phase to its capability (test-case-extractor and the per-tier test-generators for cases, quality-gate for execution, test-result-analyzer for analysis, test-code-adapter for the fix) and looping until an explicit exit condition holds. Enforces the cycle's integrity rules: a regression case before fixing a defect, a flake quarantined not retried-until-green, and the no-cheating invariant (never weaken/skip a test to force a pass). Invoke when the user asks to run the test cycle for a feature, drive a feature to green, or iterate determine-execute-analyse-fix. Don't use to run the gate once (use quality-gate), to scaffold tests without the loop (use a tier generator), or to classify results without acting (use test-result-analyzer). Supports resume per `spec/claude/resumable-work/`.
test-pyramid-check
Audits a feature's or module's test-tier completeness against the closed functional-tier taxonomy in spec/project/test-pyramid-foundation/ (are the applicable tiers — unit, component, integration, contract, E2E — present and written at the lowest tier that gives confidence?) and whether the E2E tier follows the disciplines in spec/project/e2e-test-automation/ (page-object encapsulation, condition-based waits, screenshot checkpoints, markers, TC-ID traceability). Detects the stack, globs the test files per tier, and returns a gap report. Invoke when the user asks to "check the test pyramid," "audit test-tier completeness," "verify all test levels exist," after a feature is implemented, or before a release; also handles equivalent German-language requests. Don't use to scaffold E2E tests (use e2e-test-generator), to review/repair an E2E suite (use e2e-test-reviewer), to review a run's screenshots (use e2e-result-reviewer), or to run the lint/typecheck/test gate (use quality-gate).
webview-ui-optimize
Audits a browser-rendered frontend against the canonical-language file under spec/frontend/webview-ui-optimization/ and, with per-item user approval, patches findings across five domains: Performance, Security, Accessibility (WCAG 2.2 AA), Internationalisation, and UX. Three operations: `audit` (written to `.audits/webview-ui-optimize/`), `patch` (one finding at a time), `expert-review` (dispatches `webview-ui-expert`). Invoke when the user asks to "audit the frontend", "check the UI against the webview-ui spec", "optimise for performance / a11y / i18n / UX / security", "fix the CSP", "wire up vitest-axe", or equivalent German-language requests. Don't use for brand-design decisions, audience artefact (`audience-identify`), prose linting (`prose-vale-curator`), CVE audits (`dependency-audit`), or the release pipeline (`release-publish-trigger`). Supports resume on re-invocation per `spec/claude/resumable-work/`.
gemini-image-handoff
Runs a semi-automatic Gemini image-generation handoff. The automated half authors a Gemini-optimised prompt from a brief per `spec/design/gemini-image-generation/`; the manual half guides the operator to paste it into the Gemini web UI (the Gemini app or AI Studio) and download the image from the chat. No API call, no billing, no `GEMINI_API_KEY`, and the skill writes no image and no sidecar — the operator places the downloaded file. Invoke when the user wants a Gemini image without enabling API billing, asks for a "Gemini prompt to paste into the UI", wants the manual Gemini chat route, or makes an equivalent German-language request ("Bild über die Gemini-UI generieren", "Prompt zum Einfügen in Gemini"). Don't use to call the Gemini API and write a file (use image-generate --provider gemini), to author a brand-conformant prompt document (use graphic-prompt-generator), or for FLUX/Cloudflare generation (use image-generate). Resume is not applicable: the handoff is a single interactive turn.
image-generate
Generates an image from a text prompt via a pluggable provider backend, writing the image plus a `<image>.meta.json` sidecar to an operator-chosen path. Backends are swappable via `--provider`: cloudflare (Cloudflare Workers AI FLUX.1-schnell, real free tier, DEFAULT), pollinations (auth-free, but public-feed/undocumented-licence — the tool forces private=true and shows a disclaimer), gemini (gemini-2.5-flash-image, requires billing). Wraps the bundled, stdlib-only `scripts/image_generate.py`. Invoke when the user asks to "generate an image", "create a hero image or icon from a prompt", "render this prompt to a PNG", "turn a graphic-prompt-generator document into an image", or equivalent German-language requests. Don't use for image editing, in-painting, or multi-turn refinement; for batch pipelines; or to author the prompt itself (use graphic-prompt-generator). Supports resume is not applicable: a generation is a single terminal call.
agent-review
Reviews a Claude Code agent against spec/claude/agent-management/ and spec/claude/skill-vs-agent/, and emit an actionable review plan per spec/claude/review-plan/ under .audits/agent-review/ keyed by the target agent's name. Invoke when the user asks "review this agent", "audit a specific agent file", "check whether this agent is spec-compliant", or "agent review for a specific agent". Also handles closing an existing review plan once every item is addressed — "close the agent review plan for a specific agent". Also handles equivalent German-language requests. Do NOT use for skill review (use skill-review) or for pull-request-level review (`review` skill). Supports resume on re-invocation per `spec/claude/resumable-work/`.
audience-identify
Runs the audience-identification methodology from spec/project/audience-identification/ against a bounded context (software module, service, library, or whole project) and produces an authoritative audience artifact. Invoke when the user says things like "identify audiences for this module", "who are the consumers/users of X", "stakeholder identification for X", "audience analysis before I write the README / SLA / threat model", "list the audiences of this library", or equivalent German-language requests. Also triggers when another spec or skill (readme-structure, future SLA or threat-modeling specs) needs to reference an audience list that does not yet exist for the current context. Don't use to review an existing audience artifact for compliance (use audience-review). Supports resume on re-invocation per `spec/claude/resumable-work/`.
backstage-catalog-generate
Generates a schema-valid Backstage catalog-info.yaml from an existing software project, conforming to spec/project/backstage-catalog-generation/. Inspects the repo (language/structure, remote slug, CODEOWNERS, OpenAPI/AsyncAPI/GraphQL/gRPC specs, colocated docs), infers the per-kind MUST-floor fields, confirms what it cannot infer (owner, spec.system, dependsOn, lifecycle) with the operator, writes the descriptor at the repo root, and self-validates. Invoke when the user asks to "generate a catalog-info.yaml", "onboard this repo into Backstage", "create a Backstage Software Catalog entity", "add this service to our developer portal", or German equivalents ("erzeuge eine catalog-info.yaml", "nimm dieses Repo in Backstage auf"). Optionally emits a Tech Radar TechRadarLoaderResponse JSON (never a catalog entity). Don't use to configure a Backstage backend / catalog.providers, install the Tech Radar UI plugin, or ingest Group/User org entities. Single-shot; resume not applicable.
blog-author-trigger
Operationalises spec/project/blog-author-trigger/: on a feature→done transition it derives a blog-post briefing from the feature record, computes a new/update/defer suggestion from the blog consumer's existing-post index, presents the three-way operator choice, and either dispatches blog-author with the derived briefing (Choices 1 and 2, via the cross-repository handover with explicit operator confirmation) or writes a deferral artefact under project/blog-triggers/ (Choice 3). Typically automatically dispatched by sprint-execute right after a feature reaches done; also invocable directly ("trigger a blog post for feature X", "Blog-Trigger für Feature X", "soll ich zu Feature X bloggen?"). Don't use to draft the post itself (use blog-author), to define the feature record (feature spec), to close a sprint (sprint-review), or to publish a release (release-publish-trigger). Supports resume on re-invocation per spec/claude/resumable-work/.
blog-author
Drafts a bilingual EN-canonical + DE-translated blog-post pair per this plugin's blog-author, post-writing-style, and post-audience-communication specs. Walks the operator through briefing inputs (topic-as-thesis, grounded artefact, primary audience, source list, slug, cross-language binding key), writes the EN draft, runs the pre-handover self-check, writes the DE translation, runs the per-pair self-check, executes the consumer's build command (reference `task build`), and dispatches `lektorat-apply` for the editorial audit. Invoke when the user asks to "schreibe einen neuen Blogpost", "draft a blog post about X", "neuer Eintrag zu Y", "operationalize blog-author", or equivalent German-language requests. Don't use to lektor an existing post (use `lektorat-apply`), to author portfolio pages (those live in the consumer's portfolio collection), or to redefine the consumer's post-frontmatter schema. Supports resume on re-invocation per `spec/claude/resumable-work/`.
continuous-improvement-triage
Operationalises spec/project/continuous-improvement/ by triaging portfolio audit findings, classifying them against the available specialist catalog, and dispatching hands-on remediation to the most specialised available Claude agent or skill. Invoke when the user asks to "triage continuous-improvement findings", "classify a portfolio-improvement opportunity", "dispatch findings to specialised agents", "run the quarterly specialist-coverage review", "check whether a finding class needs a new specialist", or equivalent German-language requests. Drives the three-operation loop: audit (periodic specialist-coverage review), update (record decisions and dispatch specialists), close (terminate the triage cycle). Applies the three-recurrence gap-closure rule and records all dispatch decisions in fix-PR Risk / rollout notes. Do not use to perform the hands-on remediation itself—that belongs to the dispatched specialist. Supports resume on re-invocation per `spec/claude/resumable-work/`.
cookiecutter-template-manage
Manages the lifecycle of a Cookiecutter template: scaffolds a new template or refactors an existing one (with mid-flow user confirmation of variable names and choice defaults), hardens Cookiecutter hooks, and sets up a pytest-cookies test harness. Invoke when the user says "manage a Cookiecutter template lifecycle", "scaffold or refactor a Cookiecutter template (with name + purpose confirmation)", "harden cookiecutter hooks", "set up pytest-cookies harness", "Cookiecutter-Template anlegen", "Cookiecutter-Template überarbeiten", "Cookiecutter-Hook absichern", or "pytest-cookies einrichten". Don't use for plain template consumption (a bare `cookiecutter <url>` call needs no orchestration), generic Python bootstrap, Copier or cruft work, or intentionally diverging templates. Supports resume on re-invocation per `spec/claude/resumable-work/`.
docs-audience-tracks-apply
Audits a repository against the canonical-language file under spec/project/docs-audience-tracks/ and, with per-item user approval, scaffolds or patches the documentation-tracks layer: per-page `track:` frontmatter across the per-language docs/ tree, required user-docs and developer-docs content blocks, and the audience-to-track mapping in the project's audience artefact. Three operations: `audit` (read-only conformance report), `migrate` (brownfield), `patch` (additive fixes one finding at a time). Invoke when the user asks to apply, audit, migrate, or patch documentation tracks against the spec; also handles equivalent German-language requests. Don't use for the MkDocs skeleton (`mkdocs-structure-apply`), audience artefact (`audience-identify`), page content (`audience-doc-author`), drift detection (`docs-freshness-checker`), or prose mechanics (`prose-vale-curator`). Supports resume on re-invocation per `spec/claude/resumable-work/`.
docs-dry-refactor
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/`.
feature-decompose
Decomposes a roadmap item into feature files under project/features/ per spec/project/feature/. Invoke when the user asks to decompose a roadmap item, break down a roadmap entry into features, draft features, scaffold a feature file, plan features for the next sprint, or write a new feature. Also handles equivalent German-language requests. Walks the operator through title, description, 3-7 testable acceptance criteria, and test hooks per feature; identifies which carries `verifies_sprint_value`; dispatches `feature-consistency-reviewer` (or records a manual fallback) before allowing the feature to leave `draft`. Don't use to transition feature status (`ready → in_progress` / `in_progress → done` is `sprint-execute` / `sprint-review`) or to author roadmap items, sprints, or the mission file. Supports resume on re-invocation per `spec/claude/resumable-work/`.
github-issue-templates-apply
Applies the canonical-language file under spec/project/github-issue-templates/ to a target repository — detect the project type, resolve or dispatch the audience artefact, derive triage questions, and scaffold or update .github/ISSUE_TEMPLATE/ (bug_report.yml, feature_request.yml, config.yml, plus project-type-specific extras) as GitHub Issue Forms. Invoke when the user asks to "generate issue templates for this repo", "scaffold GitHub issue forms", "create bug and feature templates", "set up .github/ISSUE_TEMPLATE", "apply the github-issue-templates spec", or equivalent German-language requests. Don't use for pull-request templates (that's `pull-request-workflow`), CODEOWNERS / SECURITY.md, discussion templates, or generic .github/ scaffolding (that's `project-structure-apply`). Supports resume on re-invocation per `spec/claude/resumable-work/`.
issue-orchestrate
Orchestrates a raw GitHub issue to an open, audit-trailed pull request per `spec/project/issue-orchestration/`. Comprehends the issue (body, comments, labels, linked items, repo surface), classifies it (`bug / feature-request / spec-change / security / docs / refactor / question / infra`), decomposes it into atomic specialist-ready work packages persisted as a pre-analysis artifact, routes large issues into the formal roadmap→feature→sprint pipeline, dispatches each package to the most specialised available skill or agent resolved by runtime lookup, and verifies via `quality-gate` and the standard PR flow. Invoke when the user asks to "analyse this issue", "orchestrate issue #N", "take this issue end-to-end", or equivalent German requests. Don't use to merge the PR (use `pull-request-merge`), to triage a red CI run (use `workflow-health-triage`), or to decompose an existing roadmap item (use `feature-decompose`). Supports resume per `spec/claude/resumable-work/`.
lektorat-apply
Reviews existing Markdown prose against six editorial dimensions (readability, comprehensibility, grammar, style, audience-fit, idiomatic naturalness) defined in `spec/project/lektorat/`. Three operations — `audit` (read-only report), `patch` (one finding, one diff, one approval), `revise` (full-artefact rewrite with diff review); dispatches `lektorat-scanner` for the detection phase. Invoke when the user asks to "lektoriere README.md", "audit docs for audience-fit", "revise this page with Lektorat", "prüfe auf Lesbarkeit", or equivalent EN/DE requests. Writes outputs to `.audits/lektorat/<YYYY-MM-DD-HHMM>/`. Don't use to author new prose (use `audience-doc-author`), curate Vale rules (use `prose-vale-curator`), lektor `spec/` files (out of scope), or edit code, configs, or LLM-instruction artefacts (SKILL.md, agents/*.md). Supports resume on re-invocation per `spec/claude/resumable-work/`.
lektorat-auto-revise
Autonomously works off a `lektorat audit` findings report by routing each artefact by type — documentation to the `audience-doc-author` agent (fully autonomous), blog posts to the `blog-author` skill (assisted) — composing a per-file briefing that binds the audience artefact and writing-style spec, letting the author revise, then re-auditing until no finding above the severity floor remains and there is no regression. Operationalises `spec/project/lektorat-auto-revise/`. Invoke when the user asks to "arbeite das Lektorat-Audit automatisch ab", "auto-revise the audit findings", "remediate the lektorat findings with the right author", or equivalent EN/DE requests. Don't use for the read-only audit or interactive per-finding fixes (use `lektorat-apply`), or for first authorship (`audience-doc-author`/`blog-author`). Writes to `.audits/lektorat-auto-revise/<YYYY-MM-DD-HHMM>/`. Supports resume per `spec/claude/resumable-work/`.
mermaid-diagrams-apply
Audits and applies the MkDocs Mermaid setup of the current repository against `spec/project/mermaid-diagrams/<canonical_language>.md`, and helps an author add a single Mermaid diagram (hand-described or derived). Wires up `pymdownx.superfences`, pins `pymdown-extensions`, refuses `mkdocs-mermaid2-plugin`. Picks the diagram type from the supported catalog and prepends the mandatory `<!-- diagram-source ... -->` marker. Audits flag missing source markers, missing direction headers, inline styling, `gitGraph` usage, non-English labels, and derived-source drift. Invoke when the user asks to "wire up Mermaid", "audit Mermaid setup", "draft a flowchart", or equivalent German-language requests. Don't use for general MkDocs scaffolding (use `mkdocs-structure-apply`), spec authoring (use `spec`), the docs-freshness audit (use `docs-freshness-checker`), or non-Mermaid diagrams. Supports resume on re-invocation per `spec/claude/resumable-work/`.
mission-define
Authors a project's first `project/mission.md` per the canonical-language file under spec/project/mission/. Invoke when the user says "define the mission", "write project/mission.md", "set up the mission file", "draft the SMART mission", or equivalent German-language requests. Walks SMART one letter at a time (Specific statement, Measurable verifies_via pointer, Achievable MVP scope, Relevant outcome IDs, Time-bound shape), gathers a per-audience MVP-deliverable paragraph, composes the frontmatter plus the four required body sections (Statement, Audiences, Verification, Source), and writes the file with `mvp_status: defining`. Refuses to run when `project/mission.md` already exists (use `mission-revise`) or when `project/goals.md` or the audience artefact is missing. Supports resume on re-invocation per `spec/claude/resumable-work/`.
mission-revise
Revises an existing `project/mission.md` per the canonical-language file under spec/project/mission/. Invoke when the user says "revise the mission", "update project/mission.md", "flip mvp_status", "the MVP is achieved", "the stabilisation gate is satisfied", or equivalent German-language requests. Supports three operations: (1) revise statement, audiences, verifying-feature pointer, or time_bound; (2) flip `mvp_status` along the legal lifecycle (`defining→in_progress→achieved→stabilised`, plus regression path); (3) revise after stabilisation with the mandatory rationale. Verifies the stabilisation-gate conditions by reading `project/roadmap.md` and `project/sprints/` before allowing a flip to `stabilised`. Supports resume on re-invocation per `spec/claude/resumable-work/`.
mkdocs-structure-apply
Audits a repository against the canonical-language file under spec/project/mkdocs-structure/ and, with per-item user approval, scaffolds or patches the MkDocs skeleton: the per-language docs/ tree, seven standard nav sections, plugin baseline (incl. mkdocs-include-markdown-plugin), required `site_url`, pinned dep manifest, per-page frontmatter contract. Three operations: `audit` (read-only conformance report), `scaffold` (greenfield), `patch` (additive fixes). Invoke when the user asks to apply, audit, scaffold, or patch MkDocs against the spec; also handles equivalent German-language requests. Don't use for theme/typography decisions (per-repo), page-content authoring (use `audience-doc-author`), DRY refactoring (use `docs-dry-refactor`), per-page track frontmatter or audience-track content blocks (use `docs-audience-tracks-apply`), drift detection (use `docs-freshness-checker`), or catalog generator wiring (use `skill-agent-catalog-apply`). Supports resume on re-invocation per `spec/claude/resumable-work/`.
permission-allowlist-maintain
Curates the committed `.claude/settings.json` `permissions.allow` list of the current repository per `spec/claude/permission-allowlist/`. Proposes additions sourced from the `fewer-permission-prompts` built-in or from the user, applies the spec's three-condition selection criteria (frequent + read-only + not-already-autoallowed), rejects forbidden pattern classes (interpreter wildcards, task-runner wildcards, mutation-capable `gh`/`git` wildcards), and routes every change through the standard pull-request flow. Invoke when the user asks to "tidy the permission allowlist", "add `Bash(task lint)` to the allowlist", "review `.claude/settings.json`", or equivalent German-language requests. Don't use to edit `.claude/settings.local.json` or `~/.claude/settings.json` (out of scope per spec). Supports resume on re-invocation per `spec/claude/resumable-work/`.
portfolio-audit
Audits, renders, and bootstraps the cross-repository capability portfolio across `nolte/*` per `spec/portfolio/portfolio-management/`. Audit dispatches portfolio-manifest-collector agent for read-only inventory collection, then detects capability duplicates, surfaces gaps, verifies tech-stack consistency per `spec/portfolio/tech-stack/`, and writes a Findings-Report under `.audits/portfolio/` using Critical / Warning / Suggestion / Info severities. Render regenerates the aggregated inventory under the per-language docs/ portfolio subtree. Bootstrap creates a repository's first `project/portfolio.yml`. Invoke when the user asks to "audit the portfolio", "check for portfolio duplicates", "render the portfolio inventory", "discover the tech stack" (audit view), or equivalent German-language requests. Don't use to consolidate duplicates (operator opens cross-repo PRs), to author new capabilities, or for per-repo tech_stack capture or refresh (use tech-stack-capture). Supports resume on re-invocation.
portfolio-inflight-triage
Runs the read-only periodic in-flight audit across nolte/* per `spec/portfolio/portfolio-inflight-management/`. Dispatches `portfolio-inflight-collector` for the four data sources (open issues, open PRs incl. drafts, branches without active PR, unresolved review threads + Discussions), applies spec stalling thresholds with optional `project/inflight.yml` overrides, classifies findings via the four-axis matrix into `Critical`/`Warning`/`Suggestion`/`Info`, attaches a specialist slug per finding, and writes a dated Findings-Report under `.audits/portfolio-inflight/`. Invoke when the user asks to "audit the portfolio in-flight", "run the in-flight triage", "check stalled PRs / issues / branches", or equivalent German-language requests. Don't use to merge / close / delete / resolve anything (operator dispatches), for per-repo CI triage (use `workflow-health-triage`), or for capability allocation (`portfolio-audit`). Supports resume on re-invocation per `spec/claude/resumable-work/`.
project-structure-apply
Audits a repository against the canonical-language file under spec/project/project-structure/ and scaffolds or patches missing artefacts: README (delegated mid-flow to `readme-structure-apply`), top-level orientation file, .gitignore, .pre-commit-config.yaml, Renovate config, Taskfile, the MkDocs site skeleton (delegated mid-flow to `mkdocs-structure-apply`), .claude/ directory, and the full .github/ layout (workflows, settings.yml, release-drafter.yml, boring-cyborg.yml, stale.yml) with the portfolio-wide Probot extends pointers. Verifies via the GitHub API that the backing GitHub Apps (Probot apps `settings`, `boring-cyborg`, `stale`, plus Renovate) are installed; for Renovate also points at the Mend dashboard when the App is installed but no activity is visible. Invoke when the user asks to audit project structure, scaffold missing GitHub configs, generate release-drafter config, check Probot/Renovate app installation, or equivalent German-language requests. Supports resume on re-invocation.
pull-request-create
Creates a GitHub pull request that conforms to the repository's pull-request-workflow spec. Invoke when the user asks to open a PR, create a pull request, draft a PR description, create a merge request, or push the branch and open a PR. Also handles equivalent German-language requests. Don't use to merge or land an already-open PR (that's `pull-request-merge`). Verifies the feature branch is synchronized with develop, composes a Conventional-Commits title and the five-section body (Summary, Changes, Linked issues, Testing, Risk / rollout notes), autolinks any touched spec files under spec/, confirms with the user, and runs the GitHub CLI PR creation command. Supports resume on re-invocation per `spec/claude/resumable-work/`.
pull-request-merge
Promotes an open draft pull request on the current branch to a merged state on `develop`, applying repository-declared labels and passing every gate from the pull-request-workflow spec. Invoke when the user asks to promote the draft PR, ship the PR, merge the draft, or bring the PR over the finish line. Also handles equivalent German-language requests. Delegates pre-merge review to the `review` skill (and `security-review` when the diff touches security-sensitive paths), derives labels from the Conventional-Commits type and touched paths, flips draft → ready, triggers automerge by applying the `automerge` label so the repository's automerge workflow squash-merges the PR once every required check is green, and verifies the merge commit landed on `develop`. Supports resume on re-invocation per `spec/claude/resumable-work/`.
readme-structure-apply
Audits the repository's `README.md` against the canonical-language file under spec/project/readme-structure/ and, with per-item user approval, scaffolds or patches the file: H1 + tagline, CI badges, six required sections in order (Purpose, Usage, Structure, Related repositories, Status, License), the consumer-first ordering rule, the ≤200-line length budget, and link rules. Three operations: `audit` (read-only conformance report), `scaffold` (greenfield), `patch` (additive fix). Invoke when the user asks to apply, audit, scaffold, or patch the README against the spec; also handles equivalent German-language requests. Don't use for docs/ page content (`audience-doc-author`), MkDocs nav (`mkdocs-structure-apply`), Vale prose linting (`prose-vale-curator`), or the audience artefact (`audience-identify`). Supports resume on re-invocation per `spec/claude/resumable-work/`.
release-notes-curate
Augments the open release-drafter draft on develop with project-context-aware sections per the canonical-language file under spec/project/release-skill-layer/ §"Skill A — Draft notes curation". Reads the project's audience artefact, derives a section bundle from the detected project type, wraps the augmentation in stable HTML-comment markers so re-runs update in place, and writes the body back via `gh release edit` for a release tag. Invoke when the user asks to "curate the release notes", "shape the release notes for this repo", or equivalent German-language requests. Don't use to publish the release (use `release-publish-trigger`), to identify audiences (use `audience-identify`), to draft notes from scratch (use the `audience-doc-author` agent), or to scaffold issue / PR templates (use `github-issue-templates-apply`). Supports resume on re-invocation per `spec/claude/resumable-work/`.
release-publish-trigger
Validates every release-automation pre-publish gate locally, then dispatches release-publish.yml via `gh workflow run` for the open release-drafter draft on develop, per the canonical-language file under spec/project/release-skill-layer/ §"Skill B — Release publish trigger". Verifies that exactly one open draft exists, the draft tag is reachable from the develop tip, version-bearing files align under their declared transform, every required status check on develop is SUCCESS, and `.github/workflows/release-publish.yml` exists. Refuses to dispatch on any failed gate; routes red checks to workflow-health triage. Never calls `gh release edit --draft=false` directly. Invoke when the user asks to "publish the release", "trigger release publish", "ship the release", or equivalent German-language requests. Typically called by sprint-review's opt-in chain, not directly after sprint closure.
requirements-elicit
Runs the requirements-elicitation interview methodology from spec/project/requirements-elicitation/ to capture a user's requirements precisely, assuming the user often does not know exactly what they want, expresses it imperfectly, or is misread. Drives a one-question-at-a-time funnel interview, maintains a per-dimension confidence score plus a gap matrix, asks confidence-gated clarifying questions only where understanding is weak, and writes an authoritative artifact to project/requirements/<slug>.md. Invoke when the user says things like "elicit the requirements for X" or "I want to build something but I'm not sure what exactly", or equivalent German-language requests. Also triggers as the upstream gate of roadmap-plan, feature-decompose, and issue-orchestrate when a requirement artifact is missing or below threshold. Don't use to decompose an existing requirement set (use feature-decompose) or author the downstream spec (use spec). Supports resume per `spec/claude/resumable-work/`.
roadmap-init
Scaffolds the project planning pair `project/goals.md` and `project/roadmap.md` for the first time, per `spec/project/roadmap/`. Invoke when the user asks to "set up the roadmap", "initialise project goals", "create goals.md and roadmap.md", "bootstrap the roadmap", or any equivalent fresh-bootstrap request. Also handles equivalent German-language requests. Verifies the audience artefact exists (and dispatches `audience-identify` when it doesn't), drafts the Vision plus numbered Outcomes in `project/goals.md`, drafts an empty queue plus optional phase headings in `project/roadmap.md`, presents both files for explicit approval, then writes them. Do NOT use to add roadmap items, retarget sprints, or flip MVP flags — that is `roadmap-plan`. Do NOT use to enforce the detail-level invariant — that is `roadmap-refine`. Supports resume on re-invocation per `spec/claude/resumable-work/`.
roadmap-plan
Adds, retargets, and reshapes roadmap items in `project/roadmap.md` per `spec/project/roadmap/` and `spec/project/mission/`. Invoke when the user asks to "add a roadmap item", "queue work for sprint N", "promote roadmap item to fine", "retarget R-3 to sprint 9", "flip MVP on this item", or equivalent German-language requests. Validates outcome IDs against `goals.md`, validates `target_sprint` against `project/sprints/`, enforces the lifecycle transitions declared in the roadmap and mission specs (including the asymmetric MVP-flag rule: `false→true` allowed before stabilisation, `true→false` forbidden after the item entered `status: active`), and refuses partial writes that would leave inconsistencies. Don't use to scaffold the roadmap from scratch (use `roadmap-init`) or to enforce the detail-level invariant (use `roadmap-refine`). Supports resume on re-invocation per `spec/claude/resumable-work/`.
roadmap-refine
Enforces the detail-level invariant on `project/roadmap.md` per `spec/project/roadmap/` §Detail-level convention. Invoke when the user asks to "refine the roadmap", "check roadmap detail levels", "is the roadmap ready for the next sprint", "promote roadmap items to fine", or equivalent German-language requests. Resolves the current and next sprint by reading `project/sprints/`, walks every roadmap item, emits a structured violation record on stderr for every item with `target_sprint` equal to the current or next sprint and `detail` other than `fine`, exits non-zero when any violation is open, and walks per-item fix proposals one at a time. Don't use to add items, retarget sprints, or flip MVP flags (use `roadmap-plan`); don't use to scaffold the roadmap from scratch (use `roadmap-init`). Supports resume on re-invocation per `spec/claude/resumable-work/`.
skill-agent-catalog-apply
Wires up the MkDocs skill-and-agent catalog in the current repository per the canonical-language file under spec/claude/skill-agent-catalog/. Audits the MkDocs config against the spec, scaffolds or patches the catalog generator surface (the `on_pre_build` hook by default, or a standalone pre-build step; gen-files is forbidden under mkdocs-static-i18n folder mode) plus literate-nav, writes the generator module that walks every configured plugin source root, and verifies a docs build produces Skills and Agents sections. Invoke when the user asks to "apply the skill-agent-catalog spec", "wire up the catalog generator", "scaffold the skills/agents navigation", or "add another plugin source root". Also handles equivalent German-language requests and checking whether a wired catalog is still in sync. Don't use for authoring individual skills/agents (use `skill-management`) or for general docs scaffolding (use `project-structure-apply`). Supports resume on re-invocation per `spec/claude/resumable-work/`.
skill-review
Reviews a Claude Code skill against spec/claude/skill-management/ and spec/claude/skill-vs-agent/, and emit an actionable review plan per spec/claude/review-plan/ under .audits/skill-review/ keyed by the target skill's name. Invoke when the user asks "review this skill", "audit a specific skill folder", "check whether this skill is spec-compliant", or "skill review for a specific skill". Also handles closing an existing review plan once every item is addressed — "close the skill review plan for a specific skill". Also handles equivalent German-language requests. Do NOT use for agent review (use `agent-review`) or for pull-request-level review (`review` skill). Supports resume on re-invocation per `spec/claude/resumable-work/`.
skills-agents-sweep
Orchestrates a portfolio-wide audit of all skills and agents in the plugin inventory, producing a consolidated sweep report under .audits/skills-agents-sweep/ with cross-cutting findings (boundary conflicts, spec-induced gaps, operations-vocabulary drift, classification errors) and a wave-based implementation roadmap. Invoke when the user asks to "run a portfolio-wide skills and agents sweep audit", "check cross-cutting drift between skills and agents", "consolidate per-artefact reviews into a single sweep report", or "plan a wave-based implementation roadmap for sweep findings". Also handles equivalent German-language requests. Do NOT use for per-artefact reviews (use skill-review or agent-review for those); do NOT use for spec-versus-implementation reconciliation (use spec-drift-audit). Supports resume on re-invocation per `spec/claude/resumable-work/`.
spec-drift-audit
Audits every spec under spec/<topic>/<slug>/ against the repository implementation (source code, config, workflows, docs) and produces a traceable audit artifact per spec/project/spec-drift-audit/. Invoke when the user asks to "run the quarterly spec-drift audit", "check spec versus implementation", "spec drift reconciliation", "audit all specs against the repo", "find spec vs code drift", "open the next quarterly audit", or when a spec changes and requires a matching partial audit. Also handles equivalent German-language requests. Do NOT use for continuous CI health checks (use workflow-health-triage), single-spec readiness review (use spec-readiness-reviewer agent), or feature-level code drift on new features (use feature-consistency-reviewer agent within feature-decompose scope). Supports resume on re-invocation per `spec/claude/resumable-work/`.
maturity-assess
Assesses the build maturity of an application's capabilities per spec/project/capability-maturity-assessment/, writing project/maturity/<slug>.md to surface where implementation quality is lacking. The `assess` operation inventories business-facing capabilities top-down (never bottom-up from directories), maps each to an audience-identification audience, dispatches the read-only capability-maturity-scanner, then grades each on three axes — completeness, code quality, test coverage — into Bronze/Silver/Gold with a weakest-link overall tier and improvement lever; `reassess` re-grades as a diff. Consumes AUDIENCES.md under a soft gate (warns, never blocks). Grading only — never a merge gate, prioritisation, or dashboard. Invoke to assess capability maturity, grade an app's functions, or produce the maturity matrix; also German. Don't use for the binary PR gate (quality-gate), to define audiences (audience-identify), or KPIs (kpi-derive). Supports resume per spec/claude/resumable-work/.
source-code-review
Runs a holistic senior-engineer source-code review of production and test code against spec/project/source-code-review/ and persists a severity-classified report whose disjoint work packages are parallel-dispatchable to specialists. The default `review` operation detects the target language and dispatches the matching read-only language reviewer agent (`python-code-reviewer`; a language without a profile is reported unsupported), then persists the report under .audits/source-code-review/. The `plan` operation dispatches implementation-plan-author with the report so specialists remediate the work packages in parallel. Invoke to review a codebase like an experienced developer, find domain-knowledge duplication, or review test-code quality; also German. Don't use for the deep OWASP audit (code-security-reviewer), CVE scanning (dependency-audit), single-tier test conformance (tier reviewers), or to run lint and tests (quality-gate). Supports resume.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.