Hermes ACP Orchestrator

VerifiedCaution

Use when you need ACP-style delegation in Hermes with explicit agent routing to hermes, codex, or claude-code, including safe output limits and timeout-aware execution.

Sby Skills Guide Bot
DevelopmentIntermediate
7/23/2026
Claude CodeCodex
#acp#delegation#multi-agent#orchestration#codex

Recommended for

Automated summary

This skill applies an ACP-style delegation pattern in Hermes to explicitly route tasks to hermes, codex, or claude-code, with bounded output size and timeout-aware execution.

Strengths

  • Explicit, deterministic routing to a chosen agent (hermes, codex, or claude-code)
  • Supports parallel batches with per-task agent overrides
  • Built-in guardrails: max_iterations, external timeout, and output character cap
  • Concrete operating rules (rich context, narrowed toolsets, structured summaries)

Limitations

  • Restricted to three delegation targets; any other agent is rejected
  • Assumes the delegate_task tool and the underlying Hermes configuration exist
  • Error handling is only covered by the brief troubleshooting section
When to use it

Use it when a task should be handed to a specific agent with isolated context and results bounded in both size and duration.

When not to use it

Avoid it when a single direct tool call is enough or the task is purely mechanical and better handled by execute_code.

Security analysis

Caution
Quality score84/100

The skill is not overtly malicious: it contains no destructive commands, secret exfiltration, obfuscated payloads, or safety-disabling instructions. However, it intentionally orchestrates external agents with terminal, file, and web access, which is powerful and could lead to unintended command execution or data exposure if used without tight scoping and approval. This warrants a caution rating rather than safe.

Findings
  • •Delegates tasks to external agent CLIs (`codex`, `claude-code`) that may execute terminal commands and access files/network.
  • •Default delegation config includes `terminal`, `file`, and `web` toolsets, increasing side-effect and data-access risk if tasks are not tightly scoped.
  • •Recommends external-agent execution but does not specify sandboxing, explicit user approval, or command allowlisting.
  • •No declared `allowed-tools`, yet examples rely on `delegate_task` and broad toolsets, creating a capability/consent mismatch.
How this analysis is made

Examples

Single targeted delegation
Delegate this bugfix to codex: implement the minimal patch, run the focused tests, and return a summary listing changed files and test results.
Mixed-agent batch
Run a batch delegation: have claude-code find regressions with the file toolset, codex implement and validate the fix with terminal and file, and hermes produce a merge-ready summary.
Timeout-safe refactor
Delegate the auth middleware refactor to claude-code with max_iterations=30, run focused tests, and report only failures.

name: hermes-acp-orchestrator description: Use when you need ACP-style delegation in Hermes with explicit agent routing to hermes, codex, or claude-code, including safe output limits and timeout-aware execution. version: 1.0.0 author: Rainhoole license: MIT metadata: hermes: tags: [acp, delegation, orchestration, codex, claude-code]

Hermes ACP Orchestrator

Overview

Use this skill when tasks should be delegated through a single ACP-style pattern while keeping context isolated and results bounded.

Targets supported by delegation:

  • hermes (default in-process subagent)
  • codex (external Codex CLI)
  • claude-code (external Claude Code CLI)

When to Use

Use this skill when:

  • You want deterministic routing to a specific agent.
  • You want mixed-agent batch delegation.
  • You need timeout and output-size guardrails for external agents.
  • You want concise summaries returned to the parent turn.

Avoid this skill when:

  • A single direct tool call is enough.
  • The task is purely mechanical and better handled by execute_code.

Delegation Patterns

1) Single task with explicit target

delegate_task(
    goal="Implement the bugfix and run tests",
    context="Keep patch minimal and include changed files in the summary.",
    agent="codex"
)

2) Parallel batch with per-task agent override

delegate_task(tasks=[
    {"goal": "Find regressions", "agent": "claude-code", "toolsets": ["file"]},
    {"goal": "Implement fix and validate", "agent": "codex", "toolsets": ["terminal", "file"]},
    {"goal": "Produce merge-ready summary", "agent": "hermes", "toolsets": ["file"]}
])

3) Reliability-focused run

delegate_task(
    goal="Refactor auth middleware",
    context="Run focused tests and report failures only.",
    agent="claude-code",
    max_iterations=30
)

Operating Rules

  1. Always provide concrete context (file paths, constraints, expected output).
  2. Keep tasks narrow; split large objectives into batch tasks.
  3. Use toolsets to reduce accidental side effects.
  4. Ask for structured summaries (what changed, what passed, what failed).
  5. For external agents, keep outputs concise and action-oriented.

Recommended Delegation Config

delegation:
  max_iterations: 50
  default_toolsets: ["terminal", "file", "web"]
  external_timeout_seconds: 900
  external_max_output_chars: 24000

Troubleshooting

  • Unsupported agent: use only hermes, codex, claude-code.
  • External agent timeout: increase delegation.external_timeout_seconds or split the task.
  • Oversized responses: lower verbosity in the delegated goal and rely on concise summaries.
  • Weak results: provide stronger context, acceptance criteria, and exact files to touch.

From the file to a first result

The download contains exactly the content displayed on this page. This copy may differ from the latest version in the source repository.

View the source and supporting files

This download contains SKILL.md only. Any scripts, references, templates and dependencies are not included. Get the required files from the source, preserving their directory structure, before continuing.

1. Download the file

Download SKILL.md

From the root of a test project, save the file as .skill-downloads/hermes-acp-orchestrator-mrx3y0l4/SKILL.md, or use the command below. This folder is for inspecting the skill before installation.

Command for macOS/Linux or WSL: requires a POSIX shell and curl. On Windows without WSL, use the download button and create the folder manually.

mkdir -p '.skill-downloads/hermes-acp-orchestrator-mrx3y0l4' && curl --fail --show-error --location 'https://skills-guides.com/api/skills/hermes-acp-orchestrator-mrx3y0l4/download' --output '.skill-downloads/hermes-acp-orchestrator-mrx3y0l4/SKILL.md.part' && mv '.skill-downloads/hermes-acp-orchestrator-mrx3y0l4/SKILL.md.part' '.skill-downloads/hermes-acp-orchestrator-mrx3y0l4/SKILL.md'

2. Check prerequisites

Read SKILL.md before using it. If it references scripts or external services, inspect them and prepare the required access in your test environment.

No recognized prerequisites section in this file. This does not mean it is self-contained: libraries, tools and access requirements may be mentioned elsewhere in the content or source repository.

The platforms on this page are catalog indications. The paths below follow tool documentation; they do not prove this skill has been run in those tools. No minimum tool version is confirmed for this listing.

3. Load the skill

After inspection, copy the complete folder into your project using one of the commands below, then open a new tool session. The command stops if a folder with the same name already exists: compare the files before replacing it.

Claude Code
test -f '.skill-downloads/hermes-acp-orchestrator-mrx3y0l4/SKILL.md' && test -s '.skill-downloads/hermes-acp-orchestrator-mrx3y0l4/SKILL.md' && mkdir -p '.claude/skills' && test ! -e '.claude/skills/hermes-acp-orchestrator' && cp -R '.skill-downloads/hermes-acp-orchestrator-mrx3y0l4' '.claude/skills/hermes-acp-orchestrator'

Codex
test -f '.skill-downloads/hermes-acp-orchestrator-mrx3y0l4/SKILL.md' && test -s '.skill-downloads/hermes-acp-orchestrator-mrx3y0l4/SKILL.md' && mkdir -p '.agents/skills' && test ! -e '.agents/skills/hermes-acp-orchestrator' && cp -R '.skill-downloads/hermes-acp-orchestrator-mrx3y0l4' '.agents/skills/hermes-acp-orchestrator'

OpenSkills

OpenSkills must already be installed. The local install/read procedure is checked with OpenSkills 1.5.0, which requires Node.js ≥ 20.6.0. Catalog slugs are not resolved automatically.

Official OpenSkills installation
openskills install './.skill-downloads/hermes-acp-orchestrator-mrx3y0l4'

openskills read 'hermes-acp-orchestrator-mrx3y0l4'

Reading should display the installed file content and its base directory. It does not run the skill or install its dependencies.

For an agent that reads AGENTS.md, then sync the index. This command changes AGENTS.md: review the diff before using it.

openskills sync

Manual method or another tool: attach SKILL.md to a conversation, or open it in an agent that can read files. Include any required supporting resources. No universal folder is recognized by every tool.

4. Try a first task

Use a short, synthetic input: an API endpoint, a small document or an interface brief. In a new conversation, attach the file or give the agent its path, then adapt this request:

Read .skill-downloads/hermes-acp-orchestrator-mrx3y0l4/SKILL.md and apply its instructions to this example: [describe a test input]. Identify any missing files, dependencies or access. Produce the result requested by the skill without inventing data or running unreviewed scripts.

Expected result: output matching the skill instructions, or a precise list of missing prerequisites. A read confirmation alone does not validate a first use. If an external resource is unavailable, that path remains untested.

Related skills