Agent Skills

Reviewed workflows for Codex, Cursor, Claude Code, and portable engineering tasks.
Browse the catalog · Open the ChatGPT plugin · Explore public skills · Contribute

Public skills are reviewed before promotion. Draft and experimental workflows stay in the incubator until they have enough evaluation proof and a clear maintenance path.

List the public catalog without installing anything:

npx skills@latest add stark-ai-de/agent-skills --list

Install one skill into the agent detected for the current project:

npx skills@latest add stark-ai-de/agent-skills --skill architecture-compass

Update installed skills:

npx skills@latest update

npx skills@latest add stark-ai-de/agent-skills --skill codex-memory-curator codex-spec-interviewer animated-readme-logo architecture-compass codegraph-ast-grep drawio-diagrams jev-capability-advisor -g -a codex -y

codex plugin marketplace add stark-ai-de/agent-skills
codex plugin add stark-ai-developer@stark-ai-developer-local

npx skills@latest add stark-ai-de/agent-skills --skill cursor-memory-curator cursor-spec-interviewer animated-readme-logo architecture-compass codegraph-ast-grep drawio-diagrams jev-capability-advisor -g -a cursor -y

npx skills@latest add stark-ai-de/agent-skills --skill claude-memory-curator claude-spec-interviewer animated-readme-logo architecture-compass codegraph-ast-grep drawio-diagrams jev-capability-advisor -g -a claude-code -y

Bundles use explicit, version-controlled manifests. The Codex bundle is the ordered skill allowlist in plugins/stark-ai-developer.source.json, and its install command below is checked against that manifest. The bundles install globally; remove -g for a project-local install. Avoid --skill '*' when targeting one runtime because it also selects skills built specifically for other runtimes.

The -a option selects the host where a skill is installed; it does not change that skill's target contract. An intentionally cross-host install preserves target-specific evidence and output while adapting collaboration controls to the selected host.

Icon	Skill	When to use
	Codex Memory Curator	Audit, review, clean up, and prune Codex memories. Use when the user asks about ~/.codex/memories, stale or noisy memories, memory pollution, cross-repo rule leakage, sensitive memory contents, memory config tuning, cleanup plans, or whether entries belong in memory, AGENTS.md, repo docs, skills, config, or deletion. Do not use for ordinary repo docs cleanup.
	Codex Spec Interviewer	Interview, source-challenge, verify, save, and ADR-gate fuzzy coding requests into Codex-ready implementation specs. Use when a feature, bugfix, refactor, migration, repo-wide change, or architecture task needs user-verified requirements, source-backed decisions, durable architecture decisions, acceptance criteria, validation commands, rollout notes, saved spec/ADR files, and a Codex execution prompt. Do not use when already fully specified or when the user wants direct implementation now.
	Animated README Logo	Audit, create, transform, or animate verified logo pipelines for GitHub READMEs. Use when a repository needs a new or reconstructed mark, motion specification, SVG animation master, executable animation recipe, static PNG, animated GIF, README-safe markup, reduced-motion fallback, or compatibility review. Do not use for unrelated app/site motion or generic image generation without a README branding target.
	Architecture Compass	Set up repository-native ADR governance, audit architecture, or plan and execute ADR-guided refactors through intent-bound workflows. Use when work needs binding agent-facing ADRs, provider-to-local mapping, architecture PR review or drift, Next.js request patterns, source placement, backend/runtime/env/config boundaries, stack deviations, or bounded ADR-governed implementation. Do not use for tiny edits, generic framework education, or work with no architecture or governance consequence.
	CodeGraph + ast-grep	Set up, update, or diagnose CodeGraph and ast-grep so coding agents can use semantic repository scope and structural syntax evidence automatically. Use when a repository needs an idempotent CodeGraph/ast-grep installation, stable tool and index migrations, MCP reconnection, persisted agent guidance, or a read-only setup diagnosis.
	Draw.io Diagrams	Create, draw, generate, edit, verify, and export draw.io/diagrams.net .drawio diagrams. Use when the user asks for editable diagrams, flowcharts, architecture, sequence, ER/UML/state, BPMN, SysML, ML/DL, swimlane, timeline, network, icon-rich technical diagrams, or PNG/SVG/PDF exports; do not use for charts/plots or artistic image generation.
	Jev Capability Advisor	Recommend available skills and tools for a supplied task, or inspect candidates locally without an API call.

The icons follow the reviewed glyph choices prepared for the stark AI Developer plugin page. See the complete public catalog for Cursor- and Claude-specific variants, exact trigger descriptions, and skill-evals/ for maintainer proof.

Find relevant available skills and tools for a supplied task, with optional local caching and inspectable recommendations. Read the usage guide or explore benchmarks and benefits, including the explicit next-skill profile: 6.3% less selection input than both measured Hussi variants in the September 25 release recheck. The current release matches all seven measured runtime files and selects a skill 6.42× as fast as the tested Native selector (GPT-6 Luna, medium); this excludes loading and task execution. Maintainers: see the release handoff for the accepted scope, package versions and publication steps.

Promotion is a folder move from incubator/skills/<category>/<skill>/ to skills/<category>/<skill>/, backed by demonstrated value, activation tests, a reusable use case, manageable maintenance, and reviewable evaluation proof. See the spec policy for public-contract updates.

pnpm install

Select checks from changed contracts and owning boundaries under ADR-0041 (Long, canonical · Guide). Run the local pnpm run validate aggregate for release intent or another mandatory gate; hosted Validate remains required for every pull request. OpenAI-native adapter packaging is proven with pnpm run validate:openai-plugin or hosted validate:release-proof, not by writing adapters/.

pnpm run scaffold repo-maintenance/my-new-skill
pnpm run scaffold:incubator engineering-workflows/my-candidate-skill

Standalone skills follow the open Agent Skills specification. The portable stark AI Developer plugin follows the Agent Plugins contract. Compatibility varies by skill, plugin client, and host; review a skill's SKILL.md and the installing client's plugin support before installation. Runtime-specific skills stay in their runtime category, while reusable workflows live under engineering-workflows or skill-maintenance.

Skills and plugins are executable context. Helper scripts are read-only unless their documentation explicitly says otherwise. Never add secrets, tokens, customer data, private repository paths, or internal hostnames to public skills, plugins, references, assets, tests, or examples.

Apache-2.0 for the skills and repository material published here.