Skip to content

Hermes Ephemeral Delegation

Lets an AI agent hand off specific tasks to fresh, short-lived worker agents for a stated reason

Stars
8,181
Licence
MIT
Files
2 files

What it does

This skill teaches an AI coding agent when it is worth splitting work across temporary helper agents instead of doing everything in one conversation. When a task has a clear reason to delegate, such as mapping a large codebase, running independent pieces of work in parallel, reviewing a diff with fresh eyes, or debugging in many steps, the agent writes a self-contained mission and starts a worker with a clean context. It then checks the worker's claims before reporting results. Small or simple work is kept inline.

Great for

  • Developers using an AI coding agent on large codebases who want to keep the main conversation's context small
  • Teams running parallel, independent changes (code, tests, docs) that can be verified separately
  • Anyone who wants an independent, fresh-context review of a risky change before merging

Skip it if

  • Small or simple tasks, where the skill itself says delegation adds cost without benefit
  • Workflows where workers must share memory across steps, since each worker starts fresh and has no memory of the parent conversation
  • Setups without a delegate_task capability in the agent runtime, because the skill depends on it

Try asking

  • “Map how authentication flows through this repository and recommend where to add rate limiting”
  • “Implement these three independent modules in parallel, each with its own tests and docs, then verify the results”
  • “Get a fresh adversarial review of this diff before I open the pull request”
  • “Debug this intermittent failure across several steps without flooding the main conversation”
Step-by-step guide: how to use Hermes Ephemeral DelegationInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 2 original files, the MIT licence and an attribution note.

    hermes-ephemeral-delegation.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip hermes-ephemeral-delegation.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip hermes-ephemeral-delegation.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • State the named reason for delegating (parallel units, high-risk verification, fresh review) so the agent does not delegate by default
  • Review the mission the agent drafts before workers run, especially the tools and paths they are allowed to use
  • Always check worker output yourself; the skill itself treats worker results as self-reports until verified

What's inside

  • references/tuning-knobs.md
  • SKILL.md

Redistributed unmodified from Gentleman-Programming/gentle-ai under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.