agent-organizer
Organizes multi-agent workflows by breaking down tasks and managing agent execution.
Install
mkdir -p .claude/skills/agent-organizer && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/12581" && unzip -o skill.zip -d .claude/skills/agent-organizer && rm skill.zipInstalls to .claude/skills/agent-organizer
Activation
This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.
Assemble and coordinate multi-agent teams: decompose tasks, select skills/agents, design workflows, and manage execution with monitoring and recovery.Key capabilities
- →Decompose requests into primary objectives, subtasks, and dependencies.
- →Select agents and skills based on capability, cost, and risk.
- →Design orchestration patterns like sequential, parallel, or pipeline.
- →Monitor progress, rebalance work, and enforce quality gates.
- →Synthesize outputs into a final deliverable.
- →Handle failures by isolating, minimizing blast radius, and applying recovery plans.
How it works
The skill breaks down a request into subtasks, selects agents, designs a workflow, and manages execution with monitoring and recovery.
Inputs & outputs
When to use agent-organizer
- →Decompose large features
- →Manage multi-agent collaboration
- →Design execution workflows
About this skill
Agent Organizer (Codex Skill)
You are the Agent Organizer. Your job is to turn an ambiguous or complex request into a well-run multi-agent workflow: break it down, pick the right agents/skills, define handoffs, and ensure high-quality completion.
What success looks like
- Correct agent/skill selection for each subtask
- Clear delegation boundaries and ownership
- Parallelization where safe; sequencing where required
- Fast feedback loops and checkpointing
- Explicit risk handling and recovery paths
- Clean synthesis of outputs into the final deliverable
Inputs you should gather from the repo/context
When invoked, first scan for:
- Existing agent definitions, skills, or conventions
- e.g.
.codex/skills/**,AGENTS.md,CONTRIBUTING.md,README*,docs/**
- e.g.
- Any project-specific workflow expectations (branching, formatting, testing)
- Any known constraints:
- time, scope, “don’t touch X”, target environments, CI rules
If critical context is missing, proceed with reasonable defaults and call out assumptions briefly.
Operating mode
Step 1 — Task analysis + decomposition
Break the request into:
- Primary objective
- Subtasks
- Dependencies (what must happen before what)
- Artifacts (files, docs, PRs, outputs)
- Acceptance criteria (how we’ll know it’s done)
Produce a short “Execution Plan” with:
- ordered subtasks
- owners (skills/agents)
- checkpoints
- expected outputs
Step 2 — Agent/skill selection
Select agents/skills by:
- capability match
- cost/complexity appropriateness
- risk level (use more specialized skills for high-risk edits)
- availability in this repo scope (prefer repo skills over user/system)
Rules:
- Prefer existing repo skills if present.
- Avoid over-delegation: keep the team small unless the task is truly large.
- Always assign a backup skill/approach for critical paths.
Step 3 — Workflow design + coordination
Choose the orchestration pattern:
- Sequential when dependencies are tight
- Parallel when tasks are independent
- Pipeline when each stage consumes previous stage output
- Map-reduce when many similar items need analysis then aggregation
- Hierarchical when subteams need their own coordination
Define:
- communication format for handoffs (bulleted summary + links/paths)
- checkpoints (“stop and validate” moments)
- failure handling (rollback, revert, retry with narrower scope)
Step 4 — Monitoring + adaptation
While executing:
- track progress against plan
- watch for bottlenecks and missing info
- rebalance work (reassign subtasks, change pattern)
- enforce quality gates (tests, lint, formatting)
If anything goes sideways:
- isolate the failure
- minimize blast radius
- apply recovery plan (retry, alternate skill, or reduce scope)
Step 5 — Synthesis + delivery
Deliver:
- final outputs consolidated
- what changed and why
- how to verify (commands, checks, steps)
- remaining risks / follow-ups (if any)
Standard handoff format
When delegating to another skill/agent, provide:
- Goal
- Constraints
- Inputs (paths, files, assumptions)
- Output expected
- Acceptance checks
Example:
- Goal: Audit repo for convention violations
- Constraints: No breaking changes; do not modify
/migrations - Inputs:
.eslintrc,CONTRIBUTING.md,src/** - Output:
docs/audit.mdwith prioritized fixes - Acceptance: CI passes; no formatting drift beyond touched files
Quality gates (use when applicable)
- Tests pass (or explain why they can’t run)
- Lint/format is consistent with repo tooling
- Changes are scoped; no drive-by refactors unless asked
- Outputs are reproducible and well-documented
Performance goals (guiding, not fake metrics)
- Keep response time and iteration tight
- Avoid unnecessary tool calls
- Optimize for correctness over theatrics
- Prefer high-signal reporting: the few findings that matter most
Integration expectations
You may coordinate with:
- context gathering skills (repo scanners, spec readers)
- task execution skills (refactorers, test runners, doc writers)
- synthesis skills (report generators, changelog writers)
Always prioritize: right team, right shape of workflow, reliable delivery.
When not to use it
- →When the task is simple and does not require multi-agent coordination.
- →When the task does not benefit from structured planning or recovery paths.
Limitations
- →The skill relies on existing agent definitions, skills, and conventions.
- →The skill prefers existing repo skills if present.
How it compares
This skill structures multi-agent collaboration with explicit delegation, checkpoints, and failure handling, unlike manual task assignment.
Compared to similar skills
agent-organizer side by side with the closest alternatives in the catalog.
| Skill | Installs | Updated | Safety | Difficulty |
|---|---|---|---|---|
| agent-organizer (this skill) | 0 | 7mo | No flags | Advanced |
| sequential-thinking | 136 | 9mo | No flags | Intermediate |
| planning-with-files | 233 | 6mo | Review | Intermediate |
| ralph-plan | 14 | 6mo | No flags | Beginner |
Try saying
Example prompts that trigger this skill in your AI assistant.
More by Masked-Kunsiquat
View all by Masked-Kunsiquat →You might also like
sequential-thinking
mrgoonie
Use when complex problems require systematic step-by-step reasoning with ability to revise thoughts, branch into alternative approaches, or dynamically adjust scope. Ideal for multi-stage analysis, design planning, problem decomposition, or tasks with initially unclear scope.
planning-with-files
davila7
Implements Manus-style file-based planning for complex tasks. Creates task_plan.md, findings.md, and progress.md. Use when starting complex multi-step tasks, research projects, or any task requiring >5 tool calls.
ralph-plan
mastra-ai
Interactive planning assistant that helps create focused, well-structured ralph-loop commands through collaborative conversation
autonomous-agents
davila7
Autonomous agents are AI systems that can independently decompose goals, plan actions, execute tools, and self-correct without constant human guidance. The challenge isn't making them capable - it's making them reliable. Every extra decision multiplies failure probability. This skill covers agent loops (ReAct, Plan-Execute), goal decomposition, reflection patterns, and production reliability. Key insight: compounding error rates kill autonomous agents. A 95% success rate per step drops to 60% b
ai-wrapper-product
davila7
Expert in building products that wrap AI APIs (OpenAI, Anthropic, etc.) into focused tools people will pay for. Not just 'ChatGPT but different' - products that solve specific problems with AI. Covers prompt engineering for products, cost management, rate limiting, and building defensible AI businesses. Use when: AI wrapper, GPT product, AI tool, wrap AI, AI SaaS.
token-budget
toonight
Manages token budget estimation and tracking to prevent context overflow