Skill Codex
Use only when the user explicitly asks to perform the task through Agent Flow. Coordinate the smallest useful workflow, verify actual execution and current results, and preserve project memory.
Agent Claude Code
AI-slop detection and cleanup subagent for text, code, UI/design, docs, copy, tests, and generated artifacts.
Agent Claude Code
Architecture subagent for implementation plans, module boundaries, dependencies, sequencing, ownership, risk, and verification criteria.
Agent Claude Code
Backend execution subagent for scoped server, API, database, queue, integration, auth, or service changes from an approved plan.
Agent Claude Code
Bun runtime subagent for Bun-based JavaScript/TypeScript projects, scripts, tests, dev servers, package management, lockfiles, and Node/npm-to-Bun migration.
Agent Claude Code
Visual asset generation subagent for approved DESIGN.md work: hero images, product mockups, illustrations, empty states, icons, and brand visuals.
Agent Claude Code
Design documentation subagent for creating and updating docs/design/DESIGN.md as the contract between concept, implementation, and verification.
Agent Claude Code
Design orchestration subagent for UI, UX, visual systems, references, Pencil, Figma, Stitch, and DESIGN.md routing.
Agent Claude Code
Documentation subagent for PRDs, specifications, task records, review sections, lessons, release notes, README work, and project documentation.
Agent Claude Code
Frontend execution subagent for scoped UI, React, styling, responsive behavior, and client-state changes from an approved plan.
Agent Claude Code
Go implementation subagent for scoped Go services, Kafka producers/consumers, CLIs, packages, concurrency, tests, modules, and idiomatic refactors from an approved plan.
Agent Claude Code
Return one structured report in your own completed native response using the shared contract in references/traceable-runs.md. Do not create a separate handoff file or repeat generated report text. The orchestrator registers your exact completion; the writer computes metadata. Keep native evidence references and report…
Agent Claude Code
Marketing and growth strategy subagent for business ideas, GTM, positioning, launch, campaigns, SEO, paid ads, growth loops, sales enablement, and promotion strategy.
Agent Claude Code
Agent Flow orchestration support subagent for routing, budget selection, subagent topology, trace hygiene, delegation packets, verification evidence, and final integration under the explicit Agent Flow invocation model.
Agent Claude Code
Pencil MCP design implementation subagent for approved DESIGN.md work in .pen files, variables, layout, screenshots, exports, and visual validation.
Agent Claude Code
Product subagent for turning an idea into scope, audience, problem, value, constraints, non-goals, PRD frame, and acceptance criteria.
Agent Claude Code
Python implementation subagent for scoped backend, CLI, automation, data processing, QA scripts, tests, PDF/RAG utilities, and dependency hygiene from an approved plan.
Agent Claude Code
QA verification subagent for requirements review before development, tests, logs, reproduction, browser or simulator checks, and readiness evidence.
Agent Claude Code
Retrieval-first LLM/RAG engineer for semantic search, chunking, embeddings, reranking, vector stores, graph databases, GraphRAG, and retrieval quality evaluation.
Agent Claude Code
Research subagent for documentation, APIs, SDKs, external sources, local examples, constraints, comparisons, and source-backed findings.
Agent Claude Code
Independent final reviewer for bugs, regressions, missing tests, plan alignment, quality, security, and release readiness.
Agent Claude Code
Senior QA subagent for blocked resolution recovery, acceptance criteria review, test design, edge cases, negative cases, and QA adequacy checks.
Agent Claude Code
Independent supervising architecture subagent for second blocked resolution recovery, architect reasoning review, root-cause classification, and final retry instruction.
Agent Claude Code
TypeScript/JavaScript implementation subagent for typed app code, Node/Bun/React modules, API clients, contracts, tests, and scoped refactors from an approved plan.
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: