claude-plugins

WCAG 2.1/2.2 compliance assessment — a11y testing strategy, remediation priorities, inclusive design. Use when the user asks to 'audit accessibility', 'assess WCAG compliance', 'evaluate a11y', 'review inclusive design', or mentions screen readers, ARIA, color contrast, keyboard navigation. [EXPLICIT]

Use when the user asks to create an agent constitution, define persistent agent identity, write agent.md, generate an agent governance spec, or add a constitution-grade agent to a multi-agent ecosystem. Produces deterministic agent.md constitutions with 22 required fields covering identity, authority, governance, quality, validation, and version control.

Create deterministic Claude Code custom subagent definitions with supported frontmatter, description-driven delegation, bounded system prompts, least-privilege tools, permission and MCP policy, model selection, and validation checks. Use when the user asks to create an agent, add a subagent, make a custom agent, define an agent definition, or build an autonomous subprocess for a specific project responsibility. [EXPLICIT]

Audita un arnes agentico existente con score determinista de 35 checks: presencia, calidad (staleness por contenido, placeholders, evidencia vacia, ciclos de dependencias), comandos stack-verified, monorepo y capa de enforcement (hooks registrados que leen stdin, Stop con max_turns, alcance estructurado, ruta unica de escritura), con anti-gaming por lineas estructuradas. Usar al pedir 'auditar arnes', 'harness audit', 'score del arnes', 'por que mi agente dice done y falla'. NO para auditar skills del toolkit (usar toolkit-hardener) ni auditorias de accesibilidad.

Crea un arnes agentico de 5 subsistemas (AGENTS.md, feature_list.json, init.sh, progress.md, session-handoff.md) mas goal contract con stop conditions, comandos stack-verified y su capa de enforcement: harness.sh como unica ruta de escritura del estado, hooks de Claude Code que leen stdin (guard de alcance, guard de bash, Stop con max_turns, contexto al arrancar) y tests/hooks.sh que los ejecuta; lo puntua con un audit determinista. Usar al pedir 'crear agent harness', 'arnes para sesiones agenticas', 'AGENTS.md inicial', 'goal contract', 'harness engineering'. NO para fuzzing harnesses (dominio de seguridad) ni para compilar DAGs (usar graph-spec-compiler).

Diagnostica la causa raiz de un arnes que deriva: reporte de drift docs-vs-realidad (claims sin entradas, entradas huerfanas, evidencia rota, comandos no verificables, placeholders residuales) via reconcile determinista. Usar al pedir 'diagnosticar arnes', 'harness drift', 'progress desactualizado', 'el arnes miente'. NO para reparar (agent-harness-repair) ni para puntuar (agent-harness-audit).

Evoluciona un arnes con lo aprendido: extrae reglas desde feedback de revision hacia AGENTS.md, captura playbooks reutilizables sin PII, y aplica condiciones de retiro a los documentos de gobierno cumplidos (trabajo terminado en pasado, pendiente en imperativo). Usar al pedir 'evolucionar arnes', 'extraer reglas', 'playbook del proyecto', 'retirar plan cumplido'. NO para fixes mecanicos (agent-harness-repair).

Extiende un arnes existente sin romperlo: features nuevas con dependencias validas, paquetes de monorepo con write boundaries declaradas, y comandos nuevos verificados contra el stack antes de documentarse. Usar al pedir 'extender arnes', 'agregar feature al harness', 'arnes monorepo', 'nuevo package con harness'. NO para crear de cero (agent-harness-creator) ni reestructurar (agent-harness-refactor).

Reestructura un arnes sin cambiar comportamiento, con disciplina deletion-only para compresion: insertions=0 verificable en el receipt, unidades borradas contadas y afirmaciones sobrevivientes intactas por construccion. Usar al pedir 'refactorizar arnes', 'comprimir AGENTS.md', 'deletion only', 'reordenar harness sin cambiar reglas'. NO para simplificar la topologia del sistema (agent-harness-simplify) ni reparar fallos (agent-harness-repair).

Repara un arnes con el patch determinista que emite el validador (check-to-fix): rellenar placeholders desde templates, degradar done sin evidencia, remover dependencias colgantes, reconciliar comandos documentados-pero-inexistentes. Diff minimo, nunca mas que lo que el audit nombro. Usar al pedir 'reparar arnes', 'harness repair', 'aplicar fixes del audit'. NO para extender funcionalidad (agent-harness-extend) ni reestructurar (agent-harness-refactor).

Prueba de vida de los frenos de un arnes agentico: por cada regla verificable inyecta la violacion en un sandbox y exige que el freno muerda (audit en rojo o gate exit distinto de cero), con caso negativo obligatorio (el arnes conforme pasa) y veredicto tri-estado MUERDE/NO-MUERDE/OMITIDA. Usar al pedir 'probar que los frenos muerden', 'selftest del arnes', 'harness selftest', 'el gate esta vivo o muerto', 'regla sin comando que la haga fallar'. NO para puntuar calidad (agent-harness-audit) ni reparar (agent-harness-repair).

Aplica la menor estructura que cumpla el objetivo: degrada graph a secuencia, secuencia a prompt, y arnes completo a subset cuando el proyecto no justifica los 5 subsistemas. Cada degradacion con justificacion honesta y evidencia de que nada verificable se pierde. Usar al pedir 'simplificar arnes', 'esto es demasiado harness', 'sobra estructura', 'downgrade a secuencia'. NO para elegir primitiva de paralelismo (agentic-parallelism-decision).

This skill should be used when the user asks 'how do I parallelize this', 'which parallel primitive should I use', 'subagent vs agent teams vs workflow', 'should I use a worktree', 'run work in the background', or needs to choose between subagents, agent view, agent teams, dynamic workflows, and git worktrees. Routes to the right primitive skill; it does not re-explain each one.

This skill should be used when the user asks to 'run Claude Code agent teams', 'orchestrate teammates', 'enable CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS', 'compare agent teams with subagents', 'assign work to Claude teammates', or 'design team lead and teammate coordination'.

This skill should be used when the user asks to 'operate Claude Code agent view', 'use claude agents', 'manage background Claude Code sessions', 'dispatch multiple Claude Code agents', 'monitor background sessions', 'peek reply attach sessions', 'troubleshoot agent view', or compare agent view with subagents, agent teams, dynamic workflows, and worktrees.

Audits existing AI system architectures against best practices — structural integrity, AI quality attributes, pattern adherence, anti-pattern detection, security compliance, and technical debt inventory. This skill should be used when the user asks to 'audit AI architecture', 'review ML system quality', 'assess AI technical debt', 'evaluate AI compliance', 'detect AI anti-patterns', 'review AI security posture', or mentions AI architecture review, AI system assessment, AI quality audit, drift monitoring audit, or AI governance review. [EXPLICIT]

Guides implementation of AI system architectures — technology selection, pipeline implementation, model serving setup, monitoring deployment, and CI/CD automation. This skill should be used when the user asks to 'implement AI architecture', 'build ML pipeline', 'set up model serving', 'deploy AI system', 'implement MLOps', 'configure drift monitoring', 'set up feature store', or mentions AI implementation plan, ML infrastructure setup, model deployment guide, RAG implementation, or agent framework setup. [EXPLICIT]

Concept of Operations (CONOPS) for AI systems — system vision, stakeholder mapping, AI-human interaction spectrum, business value assessment, success metrics, and operational modes. This skill should be used when the user asks to 'define the AI operational concept', 'map AI stakeholders', 'design AI-human interaction levels', 'assess AI business value', 'define AI success metrics', 'plan AI operational modes', or mentions CONOPS, IEEE 1362, AI autonomy levels, AI value matrix, or AI system vision. [EXPLICIT]

AI-specific design patterns and system tactics — Feature Store, Champion-Challenger, Shadow Deployment, Drift Detection, Explainability Wrapper, Canary Deployment, Bulkhead, and traditional patterns adapted for AI. This skill should be used when the user asks to 'select AI design patterns', 'apply ML patterns', 'design drift detection', 'implement feature store', 'plan shadow deployment', 'design champion-challenger', 'select availability tactics for AI', or mentions AI anti-patterns, maintainability tactics, fault recovery for models, or pattern selection for ML systems. [EXPLICIT]

AI pipeline architecture design — development pipelines, production pipelines, data stores, model registry, CI/CD for AI, and non-functional requirements. This skill should be used when the user asks to 'design AI pipelines', 'architect ML pipelines', 'select data stores for AI', 'design model registry', 'implement CI/CD for ML', 'define AI pipeline requirements', or mentions MLOps, training pipeline, inference pipeline, feature pipeline, Blue and Gold deployment, or pipeline patterns. [EXPLICIT]