Orchestrating Subagents
Use when about to spawn one or more subagents, agent teams, or a workflow — a single delegated search, a parallel fan-out, or a repeated batch — and whenever choosing which model a delegated task runs on instead of letting it inherit yours. Also use when writing or editing a .claude/agents definition, when a delegated run hit a rate limit or cost more than expected, or when a subagent came back after far more or far fewer turns than the work warranted.
Install
npx skills add PapiScholz/symphony --skill orchestrating-subagentsgit clone --depth 1 https://github.com/PapiScholz/symphony.git /tmp/symphony
mkdir -p ~/.claude/skills
cp -r /tmp/symphony/skills/orchestrating-subagents ~/.claude/skills/orchestrating-subagentsPersonal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.
From the SKILL.md
# Orchestrating Subagents ## Overview Delegation has two dials — **which model**, **how much reasoning effort** — and both inherit the orchestrator's by default. Inheriting is a decision, not a neutral default: it puts the most expensive model on the cheapest work. The orchestrator stays strong; the workers rarely need to be. ## The Dispatch Contract A one-off `Agent` call takes exactly these four. All four filled, every time: ``` subagent_type: general-purpose | Explore | Plan | <custom> model: haiku | sonnet | opus | fable description: 3-5 words prompt: self-contained; states output format and where to write it ``` **`effort` is not a parameter of the `Agent` tool.** Writing it there does nothing. Effort is set where it exists: | Where | How | |---|---| | `.claude/agents/*.md` | `effort:` in the frontmatter, alongside `model:` | | Workflow stage | `opts.effort` next to `opts.model` | | One-off `Agent` call | no field — it inherits the session. Shape the work in the prompt instead: "single pass, do not second-guess" for cheap work, "work section by section" for thorough work. | Recurring role → give it a definition file and set both dials there. ## Model × Effort | The delegated work is… | model | effort | |---|---|---| | Filling a fixed schema, classifying, labeling, renaming, summarizing one file | `haiku` | `low` | | Locating code, web research, reading with bounded judgment | `sonnet` | `medium` |
Details
- Skill name
orchestrating-subagents- Repository
- PapiScholz/symphony
- Path
skills/orchestrating-subagents/SKILL.md- License
- MIT
- Repo updated
- 3 October 2026
- Listed in
- karanb192/awesome-claude-skills
FAQ
What does the Orchestrating Subagents skill do?
Use when about to spawn one or more subagents, agent teams, or a workflow — a single delegated search, a parallel fan-out, or a repeated batch — and whenever choosing which model a delegated task runs on instead of letting it inherit yours. Also use when writing or editing a .claude/agents definition, when a delegated run hit a rate limit or cost more than expected, or when a subagent came back after far more or far fewer turns than the work warranted.
How do I install the Orchestrating Subagents skill?
skills CLI (Claude Code, Cursor, Codex and more): npx skills add PapiScholz/symphony --skill orchestrating-subagents. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.
Related skills
Reels Setup
VishiATChoudhary/reels
Use when setting up, installing or onboarding the reels studio (github.com/VishiATChoudhary/reels), when a user or agent wants to start making story reels with it, or when its setup is broken (missing whisper-cli, model, yt-dlp, ffmpeg, node deps). Clones, installs, verifies, and stars the repo (one [Y/n] prompt).
Least Cost Capable
kanibal-studio/least-cost-capable
Compare declared relative cost ranks among catalogued agent models after checking task capability, availability, authorization, workload, and cost evidence.
Session Handoff
alapha888/session-handoff-kit
Use this skill when context usage approaches capacity (>60%), before running /clear or /compact, when switching tasks, when ending a coding session across Claude Code, Cursor, or Codex, or as a mid-session diagnostic — if the agent cannot fill the handoff fields from actual state, the session has drifted. Produces a structured handoff artifact that transfers execution state to a fresh agent session without loss of context or drift.
Fresh Review
roman-s8o/workflow-pack-free
Review an artifact with an independent agent that has no access to the author's reasoning, then classify its findings.
Slop Post
useslop/claude-plugins
Post what this Claude Code session built to Slop (useslop.com) as a private draft with an automatic Build Receipt (model, tools, run time, commits), optionally as an entry in this week's Slop challenge. Use when the user runs /slop-post or asks to share, post or ship this session's work to Slop. Saves a draft only; the human publishes it from the claim link.
Aeo Check
pchewlo/aeo-check
Run aeo-check on a website's code (or a live URL) to find answer engine optimisation problems that stop ChatGPT, Claude, Perplexity, Gemini and Google AI Overviews from crawling, understanding, citing or buying from the site, then fix them. Use when asked to check or improve AEO, GEO, AI search visibility, llms.txt, robots.txt for AI crawlers, structured data, or agent readiness.
Browse the most popular Claude skills, all ai, llms & agents skills, or write your own with the skills generator.