智能体
Agent 规划、多步流程与编排
plan-mode
wcygan
Research, analyze, and create a detailed implementation plan without making any code changes. Read-only planning mode with sub-agent exploration. Use when planning implementation, designing approach, creating technical plans, or analyzing before building. Keywords: plan, design, approach, implementation plan, analyze, research, strategy, read-only planning
session-continuation
nguyenthienthanh
"Manage workflow state across sessions with handoff and resume. TOON-based state persistence."
brainstorm
wcygan
Multi-agent brainstorming for software design decisions and architecture choices. Spawns parallel agents with differentiated roles (codebase explorer, trade-off analyst, devil's advocate) to evaluate options, surface blind spots, and produce structured analysis including trade-off matrices, pre-mortems, and ADR-ready summaries. Use when making architecture decisions, choosing between approaches, or evaluating trade-offs. Keywords: brainstorm, decision, trade-off, architecture, design, evaluate, options, approach, choose, pros cons, ADR, pre-mortem, WWHTBT
learning-analyzer
nguyenthienthanh
"Analyze collected learning data from Supabase to identify success patterns, failure patterns, optimization opportunities, and agent performance trends."
competitor-analysis
wcygan
Comprehensive competitor research using a 7-agent team. Spawns parallel research agents for product, marketing, UX, and technical analysis, then sequential synthesis agents for MVP spec, go-to-market strategy, and competitive landscape. Use when evaluating a competitor, planning a new product, or scoping an MVP. Keywords: competitor, competitive analysis, market research, product research, MVP, go-to-market, GTM, teardown, product teardown
agent-team
wcygan
Launch and coordinate Claude Code agent teams for complex tasks. Use when the user wants to parallelize work across multiple agents, needs a team for review, debugging, design, or feature development, or asks to "create a team" or "spawn agents." Keywords: agent team, team, spawn, teammates, parallel agents, coordinate, collaborate, multi-agent
context-decorate
wcygan
Scan a codebase and place CLAUDE.md context files in directories that lack them. Spins up a team of subagents to parallelize the work — each agent covers a few directories and generates a succinct overview of the directory's purpose and key subdirectories. Use when onboarding to a new codebase, improving Claude context quality, annotating project structure, or scaffolding directory-level context for AI assistants. Keywords: context, CLAUDE.md, annotate, codebase, directory, onboard, decorate, scaffold context, context files.
new-agent
wcygan
Create custom Claude Code subagents. Interactive wizard that gathers requirements, generates the agent .md file with proper frontmatter and system prompt. Use when creating a new agent, adding a subagent, or making a custom agent available globally. Keywords: agent, subagent, create agent, new agent, custom agent
debug-team
wcygan
Debug complex issues using an agent team with competing hypotheses. Spawns 3-5 investigators that actively challenge each other's theories and converge on the root cause. Use for tough bugs where the root cause is unclear. Keywords: debug team, competing hypotheses, root cause, investigation, multi-agent debug, hard bug, mystery bug
pm-expert
nguyenthienthanh
"Project management expertise - agile methodology, estimation, risk management, stakeholder management"
arch-review
wcygan
RFC-style review of major technical decisions using a 5-agent debate team. Spawns tech-lead, security-auditor, performance-analyst, reliability-engineer, and devils-advocate for independent analysis followed by structured discussion. Produces decision document with pros/cons, dissenting opinions, and recommendation. Use for architecture decisions, technology selection, major refactors, design reviews. Keywords: architecture, RFC, design review, technical decision, architecture review, design decision, major change
ab-analyze-backend
ayoubben18
Analyze and document a project's backend patterns — API design, data access, service structure. Use when the user wants a backend-only architecture analysis.
write-agents-entry
nesnilnehc
Write or revise AGENTS.md per embedded output contract. Use when creating Agent entry for new projects, auditing existing AGENTS.md, or adopting the AI Cortex entry format.
sync-architecture
ayoubben18
Post-implementation documentation-sync detector. Spins up ONE read-only subagent on a completed task's diff to find everything the change introduced that the architecture docs don't yet know about — new endpoints, patterns, dependencies, domain terms, ADR-worthy decisions — routed to the exact doc each belongs in. Autonomous runs apply only append-only safe additions and defer domain/ADR reshapes; interactive runs present the proposed doc deltas to pick. Use after a task's missions are done (from create-task / start-task / start-roadmap) or standalone on a diff.
ab-resume-task
ayoubben18
Continue an existing task from its progress tracker. Use when the user wants to resume, continue, or pick up earlier task work.
enterprise-agent-deployment
openamer
'Use for enterprise agent deployment tasks. Candidate synthesized from trend radar; awaiting trial promotion.'
ab-start-task
ayoubben18
Run an existing task autonomously to completion — each remaining mission in a subagent with tdd, tracker updated per mission, a commit after every green mission. Use when the user wants to hand off a well-defined task and review commits instead of missions.
review-implementation
ayoubben18
Post-implementation review. Spins up three read-only critics on a completed task's diff — cleaner-architecture, slop-defender, reusability-inspector — that push back ONLY on real issues. Autonomous runs apply safe fixes (tests-green-gated) and write everything to review.md next to progress-tracker.md; interactive runs present findings to pick. Use after a task's missions are done (from start-task / start-roadmap / create-task) or standalone on a diff.
generate-github-workflow
nesnilnehc
"GitHub Actions YAML with embedded output contract: security-first, minimal permissions, version pinning. For CI, release, PR checks. Differs from generic templates by spec compliance and auditability."
ab-mastermind
ayoubben18
The intelligent entry point to the AB Method. Use when the user isn't sure which workflow they need, asks whether to create a goal or a task, or wants to understand how the AB Method works end to end. Routes intent to the right workflow instead of reimplementing it.
ab-analyze-frontend
ayoubben18
Analyze and document a project's frontend patterns — component structure, state management, conventions. Use when the user wants a frontend-only architecture analysis.
ab-extend-goal
ayoubben18
Extend an existing goal — broaden or add to its objective, building on top of what the earlier /goal run already implemented. Use when the user wants to expand a goal after a /goal run.
curate-skills
nesnilnehc
Evaluate, score (ASQM strict), tag, and normalize all Skills; writes agent.yaml and README per skill, detects overlaps, produces ASQM_AUDIT.md or chat summary. Use when auditing skills, after adding/changing skills, or when generating repo-level skill summaries.
ab-start-roadmap
ayoubben18
Execute a roadmap autonomously to completion — the layer above start-task. Use when a roadmap's tasks are planned and you want them run in dependency order, with independent tasks optionally in parallel. Verifies the dependency tree is planned before running; stops if a needed task has no plan. Executor only.