orchestrate-agent-work

A coordination guide for assigning limited tasks to other coding agents and combining their reports.

In plain words
What is it for?
It is for splitting independent work, defining each worker's scope and acceptance criteria, tracking outcomes, handling escalations, and closing delegated tasks.
Why use it?
It prevents overlapping work, unclear ownership, silent failures, and agents taking actions outside their assigned authority.

Skill for Claude CodeCodex

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add skills/gaelic-ghost/socket/orchestrate-agent-work
Any agent
npx skills add gaelic-ghost/socket --skill orchestrate-agent-work
Clone the repo
git clone --depth 1 https://github.com/gaelic-ghost/socket

Made for: Claude Code, Codex.

Per session 40 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 550 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00040 $0.00550
Opus 5 $0.00020 $0.00275
Sonnet 5 $0.00008 $0.00110
Haiku 4.5 $0.00004 $0.00055

Measured 2d ago against content hash 3f6684d3a1db, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

orchestrate-agent-work scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

plugins/agent-engineering-skills/skills/orchestrate-agent-work/SKILL.md · 53 lines

How it starts

The opening of the file, as written. The whole thing — 53 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Orchestrate Agent Work

Use this skill at the delegation boundary. The coordinator owns the user goal, decomposition, authority, final decision, validation, and closure; workers own only their assigned slice.

Workflow

  1. Confirm delegation is requested or required by the active workflow. Prefer a single agent when the work is sequential or one small write target.
  2. Split only independent work. Assign one coordinator and one owner per artifact, worktree, branch, validation lane, and external side effect.
  3. Give every worker a launch envelope:
    • task and coordinator identity;
    • objective, bounded scope, inputs, and source-of-truth locations;
    • allowed and forbidden actions, including write authority;
    • acceptance criteria, validation owner, and output location;
    • required terminal report and escalation conditions.
  4. State whether the coordinator waits for all results, consumes partial reports, or hands work to a durable task surface.
  5. Track each worker until it is completed, blocked, needs-decision, cancelled, or failed. A worker must not silently stop.
  6. Acknowledge each result: consume it, request one bounded follow-up, mark it superseded, or cancel it explicitly. Then synthesize and verify the result in the coordinator thread.

Worker Report Contract

Require the final report to contain the terminal state, outcome, evidence or artifacts, validation run or intentionally skipped, uncertainty, and one recommended next action. Workers escalate rather than improvise when scope, authority, ownership, evidence, or acceptance criteria change.

Boundaries

  • Keep user-facing decisions, final edits, merge, release, and final validation with the coordinator unless a user-approved disjoint ownership plan says otherwise.
  • Default to one delegation layer. Nested orchestration needs an explicit reason, bounded fan-out/depth, and host support.
  • Use coordinate-external-agents when work crosses a host, session, or durable ownership boundary.
  • Use coordinate-worktrees-and-threads when workers need repository state. When the selected host is Codex and the worktree mode, local environment, or app-owned thread association needs a decision, compose it with repository-skills:codex-gui-worktree-workflow. On Hermes or another host, keep the ownership contract host-neutral and use ordinary Git worktrees.

Read the full file on GitHub · 53 lines

Files

What ships with it

1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 53 lines · 40 tokens per session scan A 3f6684d3a1db

Subscribe to this mod's changes

orchestrate-agent-work is a skill published in the GitHub repository gaelic-ghost/socket (7 stars, last pushed 6d ago), licensed Apache-2.0. It adds 40 tokens to every session and 550 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other skills, from other repositories

xcode-disk-cleanup

Audit and safely clean Xcode-related developer storage on macOS. Use whenever a developer mentions low disk space, Xcode storage, DerivedData, simulator devices or runtimes, stale beta platforms in Xcode Settings, DeviceSupport, archives, dSYMs, CocoaPods caches, simulator dyld caches, project .build folders…

AvdLee/Xcode-Disk-Cleanup-Agent-Skill · 147 tokens

storescreens

Set up and run storescreens-cli to automate App Store screenshot capture for iOS apps: render captioned/framed App Store-ready screenshots with device bezels, markdown captions, and panoramic backgrounds; upload screenshots + per-locale metadata (name, subtitle, description, keywords, what's new, promotional text) to…

ciscoriordan/storescreens-skill · 331 tokens

poll-github-work

Poll and rank open GitHub issues, RFCs, and pull requests for maintainer work, or start one explicitly selected item. Use when a user asks what to work on, requests backlog priorities, says "poll work," wants actionable issues or pull requests, or says "start.

trycua/cua · 61 tokens

test

Run tests. Use after code changes to validate. Arguments: unit (default, no GPU), e2e (with models), filter name, or all.

soniqo/speech-swift · 34 tokens

release

Build, sign, notarize, and publish a new DroidProxy release. Use when the user wants to cut a release, merge a PR and ship it, or asks to build/notarize/publish. Covers the full pipeline: PR merge, version bump, swift build, code signing, Apple notarization, Sparkle appcast update, and GitHub release creation.

nikships/droidproxy · 80 tokens

remarc

Manage Remarc session comments and contextual feedback. Use when the user mentions Remarc sessions, comments, handoff, triage, review, addressing, resolving, status updates, or summaries.

metedata/Remarc · 41 tokens