Specialist subagent that writes the actual prompts for project subagents (researcher, coder, reviewer, tester, etc). Invoked by framework-supervisor during Fase 3 of framework construction. Takes the topology decided with the user and produces well-crafted .claude/agents/.md files following the standard template.
QA specialist subagent that audits a completed agentic framework before packaging — checks CLAUDE.md completeness, agent/skill consistency, hook safety, MCP credentials hygiene, link integrity, and applies the Swiss Cheese evaluation model (deterministic checks + state assertions + LLM-as-judge). Invoked by…
The orchestrator agent for the Creador de Frameworks Agénticos plugin. Coordinates the 6 phases of building a project's agentic framework by delegating to specialized subagents (memory-architect, agent-prompter, skill-writer, hooks-engineer, mcp-integrator, framework-reviewer) and synthesizing results. Operates on the…
Specialist subagent that designs and writes lifecycle hooks for a Claude project — PreToolUse, PostToolUse, SessionStart, SessionEnd, UserPromptSubmit. Invoked by framework-supervisor during Fase 5 when governance, auditing, sanitization, or permission gates are needed.
Specialist subagent that connects a Claude project to external systems via the Model Context Protocol (MCP) — Slack, GitHub, Google Drive, Notion, Postgres, Jira, custom APIs, etc. Invoked by framework-supervisor during Fase 5 when the project needs to read/write data in services beyond the local filesystem. Also…
Specialist subagent that designs and writes CLAUDE.md, MEMORY.md, and the entire .claude/ memory hierarchy for a project. Invoked by framework-supervisor during Fases 1 and 2. Produces the project's constitution and seed memory based on inputs from the supervisor.
Specialist subagent that drafts new skills for a project — turning recurring workflows into cargable SKILL.md modules with their references and scripts. Invoked by framework-supervisor during Fase 4. Follows the skill-forger procedure.
Use when implementing a feature, fixing a bug, or making code changes that have clear acceptance criteria. Triggers include "implementa", "agrega la función", "corrige el bug donde", "refactoriza X", "make this work", or any task with a measurable code outcome. Operates strictly within CLAUDE.md rules.
Use when a technical question, decision, or comparison needs deep investigation across docs, web, codebase, or external sources before answering. Triggers include "investiga", "compara X vs Y", "qué dice la doc oficial", "research best practices for", "find out how to", or any question that requires reading 3+ sources…
Use when code, docs, or a PR needs to be audited against project standards before merging or accepting. Triggers include "revisa", "review this PR", "audita", "check the code in X", "is this OK to merge", "review my changes". Returns categorized findings, does not modify what it reviews.
Use when designing test cases, writing new test files, or executing test suites and analyzing failures. Triggers include "escribe tests para X", "test this component", "run the tests", "why is this test failing", "increase coverage of Y", "design a test plan for". Writes tests; doesn't modify production code to force…
★not rated 2 2mo agoA74 tokens
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: