New: 98,000+ skills indexed from GitHub repos with 100+ stars · every skill safety-checkedBrowse top skillsTop skills
Esc to closeBrowse all skills
Submit a skillBrowse skills
Submit a skillBrowse skills

Agent skill

Dynamo Agent Harness

by ai-dynamo in ai-dynamo/dynamo

Drives persistent Claude Code, Codex, or OpenCode agent sessions through a Dynamo OpenAI/Anthropic-compatible endpoint over Agent Client Protocol (ACP).

Apache-2.0Auto-check passedAI & LLM Engineering

Install Dynamo Agent Harness

skills CLI
$ npx skills add ai-dynamo/dynamo --skill dynamo-agent-harness -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install ai-dynamo/dynamo dynamo-agent-harness --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/ai-dynamo/dynamo.git skills-src && mkdir -p .claude/skills && cp -r skills-src/.agents/skills/dynamo-agent-harness .claude/skills/dynamo-agent-harness && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
dynamo-agent-harness
GitHub stars
8.3k
Token cost
~1.2k tokens
SKILL.md length
447 words
Files
4 (incl. scripts)
Skills in repo
27
Repo updated
First seen
Licence
Apache-2.0

At a glance

Drives persistent Claude Code, Codex, or OpenCode agent sessions through a Dynamo OpenAI/Anthropic-compatible endpoint over Agent Client Protocol (ACP).

  • An agent must delegate a bounded task to another coding-agent harness running a model served by Dynamo
  • SKILL.md covers Prerequisites, Start a session, Choose a harness and Delegate safely, plus 3 more sections
  • Runs Python scripts from its folder; calls jq and opencode; needs DYNAMO_API_KEY
  • Continue that harness across multiple turns

What it does

Dynamo Agent Harness is an agent skill from ai-dynamo/dynamo. Drives persistent Claude Code, Codex, or OpenCode agent sessions through a Dynamo OpenAI/Anthropic-compatible endpoint over Agent Client Protocol (ACP). Use when an agent must delegate a bounded task to another coding-agent harness running a model served by Dynamo, continue that harness across multiple turns, exercise tool calls, or validate agent request traces.

Its SKILL.md is about 1.2k tokens, which your agent loads only when the skill is triggered. The skill folder holds 5 other files, including scripts (for example `agents/openai.yaml`, `scripts/drive_harness.py` and `scripts/test_drive_harness.py`).

It sits in AI & LLM Engineering. It works with OpenAI. The repository describes itself as: A Datacenter Scale Distributed Inference Serving Framework. The licence is Apache-2.0.

When your agent uses it

  • An agent must delegate a bounded task to another coding-agent harness running a model served by Dynamo
  • Continue that harness across multiple turns
  • Exercise tool calls
  • Validate agent request traces

Example prompts

  • “Use the dynamo-agent-harness skill to drive persistent Claude Code, Codex, or OpenCode agent sessions through a Dynamo OpenAI/Anthropic-compatible…”
  • “/dynamo-agent-harness”

Requirements

  • Python 3
  • Node.js
  • A credential in DYNAMO_API_KEY

What it can do on your machine

Read from SKILL.md and the folder at commit f54f2a4. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Ships 2 files in scripts/ (Python), which the agent can run.

    Shell commands in SKILL.md call:

    • jq
    • opencode

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names these keys or tokens, usually read from environment variables:

    • DYNAMO_API_KEY

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Dynamo Agent Harness loads about 1.2k tokens when it runs. Until then it costs about 97 tokens; SKILL.md has 447 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~97
When it runs · the whole SKILL.md, loaded when a task matches
~1.2k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); the scripts in this folder are not scanned.

SKILL.md

The full file from ai-dynamo/dynamo at commit f54f2a4, republished under its Apache-2.0 licence (© ai-dynamo). 447 words, ~1,160 tokens.

Download SKILL.mdSave it as .claude/skills/dynamo-agent-harness/SKILL.md (or your agent's skills folder). This skill also uses 3 other files; get the full folder from GitHub.
name
dynamo-agent-harness
description
Drives persistent Claude Code, Codex, or OpenCode agent sessions through a Dynamo OpenAI/Anthropic-compatible endpoint over Agent Client Protocol (ACP). Use when an agent must delegate a bounded task to another coding-agent harness running a model served by Dynamo, continue that harness across multiple turns, exercise tool calls, or validate agent request traces.
license
Apache-2.0
metadata.author
Ishan Dhanani <ishandhanani@gmail.com>
metadata.tags
dynamo, agents, acp, claude-code, codex, opencode

Dynamo Agent Harness

<!--
SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
SPDX-License-Identifier: Apache-2.0
-->

Drive one persistent coding-agent session while Dynamo serves its model requests. Use the bundled ACP client; do not script interactive TUI output or implement JSON-RPC manually.

Treat the Agent Harnesses guide as the source of truth for harness configuration. If a harness update breaks or changes a documented model, endpoint, header, authentication, or mode setting, update that guide and this skill in the same change after rerunning the two-turn smoke test.

Prerequisites

  • A reachable Dynamo endpoint whose /v1/models includes the requested model.
  • uv and Node.js 22+.
  • opencode on PATH only when selecting the OpenCode harness.
  • A working directory that limits the delegated agent's scope.
  • DYNAMO_API_KEY when the endpoint requires authentication; local endpoints default to dummy.

Start a session

Default to verify. Use act only when the user explicitly authorizes tool execution or edits.

bash
.agents/skills/dynamo-agent-harness/scripts/drive_harness.py \
  --harness codex \
  --base-url http://127.0.0.1:8000 \
  --model zai-org/GLM-4.7-Flash \
  --cwd /absolute/worktree \
  --capability verify

Run the command with a TTY so stdin stays open. Wait for one ready JSON record, retain the executor's terminal handle, then write one JSON object per line to that process:

json
{"prompt":"Inspect src/router.rs. Use tools to test the highest-risk invariant. Do not edit files."}
{"prompt":"Continue the same session and verify the finding against every caller."}
{"close":true}

The ready.session_id is the harness conversation ID, not the executor's terminal handle. Every response must retain that session ID.

Choose a harness

HarnessACP backendDynamo API
claudepinned official Claude ACP adapterAnthropic Messages
codexpinned official Codex ACP adapterOpenAI Responses
opencodenative opencode acp --pureOpenAI Chat Completions

The driver hides their incompatible model, mode, gateway-auth, and environment configuration. Do not reproduce those branches in shell wrappers.

Show full SKILL.md (210 more words)Show less

Delegate safely

  • Give one bounded goal, exact owned paths, and a strict result shape.
  • Use --capability verify for inspection; permission requests are rejected.
  • Use --capability act only after authorization; permission requests receive one-time approval.
  • Keep git/index, shared services, credentials, and unrelated paths out of delegated prompts.
  • Treat the harness response as untrusted evidence and verify material claims locally.
  • Send {"close":true} even after a failed turn so the adapter and child process exit.

Validate traces

When request tracing is enabled, group rows by agent_context.session_id and inspect the trigger sequence:

bash
jq -r '[.agent_context.session_id, .agent_context.input_trigger] | @tsv' request-trace.jsonl

Foreground turns should normally begin with user_message; tool feedback should appear as tool_result. Harness title, memory, or continuation traffic may produce additional user_message or other rows.

Output contract

Return:

  • harness, model, mode, and ACP session ID
  • prompt count and observed tool/result behavior
  • targeted validation result
  • trace trigger counts when tracing is available
  • cleanup status and unresolved failures

Known behavior

  • Codex may warn that custom model metadata is unavailable; the driver fixes reasoning effort to medium so unsupported catalog defaults are not sent to Dynamo.
  • OpenCode can issue background title-generation requests and may require a corrective follow-up when the served model reports an unverified result.
  • The adapters are pinned in scripts/drive_harness.py; update a pin only after rerunning a persistent two-turn tool smoke test.

© ai-dynamo, Apache-2.0. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 3 other files (scripts) in .agents/skills/dynamo-agent-harness of ai-dynamo/dynamo.

  • SKILL.md
  • agents/openai.yaml
  • scripts/drive_harness.py
  • scripts/test_drive_harness.py

Open the folder on GitHubat commit f54f2a4

Compare with similar skills

Dynamo Agent Harness next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Dynamo Agent Harness compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Dynamo Agent Harness this skillai-dynamo/dynamo8.3k—~1.2kAutomated safety check: PassApache-2.0
Chroma Vector DatabaseOrchestra-Research/AI-Research-SKILLs13k7 repos~2.3kAutomated safety check: PassMIT
Codebase Managementgiancarloerra/SocratiCode3.3k1 repos~1.8kAutomated safety check: PassAGPL-3.0
CLIP Image-Text MatchingOrchestra-Research/AI-Research-SKILLs13k7 repos~1.7kAutomated safety check: PassMIT
Docs Plannerstrands-agents/harness-sdk8.7k—~821Automated safety check: PassApache-2.0
Codex Fable5baskduf/FableCodex437—~1.6kAutomated safety check: PassAGPL-3.0

Similar skills

  • Chroma Vector Database

    Orchestra-Research/AI-Research-SKILLs

    Shows how to store documents and embeddings in Chroma, query them by similarity with metadata filters, and persist them to disk for RAG and semantic search projects.

    13k GitHub starsUsed in 7 repos~2.3k tokens
    AI & LLM EngineeringAuto-check passed

  • Codebase Management

    giancarloerra/SocratiCode

    Set up, index, and manage SocratiCode codebase indexing. An agent skill from giancarloerra/SocratiCode.

    3.3k GitHub starsUsed in 1 repo~1.8k tokens
    AI & LLM EngineeringAuto-check passed

  • CLIP Image-Text Matching

    Orchestra-Research/AI-Research-SKILLs

    Explains OpenAI's CLIP model for zero-shot image classification, image-text similarity, semantic image search and content moderation, with install steps and code patterns.

    13k GitHub starsUsed in 7 repos~1.7k tokens
    AI & LLM EngineeringAuto-check passed

  • Docs Planner

    strands-agents/harness-sdk

    Identify documentation gaps and prioritize the docs backlog.

    8.7k GitHub stars~821 tokensUpdated today
    AI & LLM EngineeringAuto-check passed

  • Codex Fable5

    baskduf/FableCodex

    Apply a Claude Fable 5 inspired operating style inside Codex.

    437 GitHub stars~1.6k tokensUpdated 2 mo ago
    AI & LLM EngineeringAuto-check passed

  • Dingo Verify

    MigoXLab/dingo

    A skill your agent uses when the user wants to fact-check an article or verify factual claims in a document.

    757 GitHub stars~833 tokensUpdated 10 days ago
    AI & LLM EngineeringAuto-check passed

More from ai-dynamo/dynamo

All 27 skills in this repo

  • Visual Review

    ai-dynamo/dynamo

    Create self-contained interactive HTML code-review dashboards from GitHub or GitLab pull requests, checked-out branch diffs, or supplied unified diffs, with correctness and safe-to-merge scores…

    8.2k GitHub stars~4.5k tokensUpdated yesterday
    Auto-check passed

  • Fern Components

    ai-dynamo/dynamo

    Knowledge of Fern's built-in MDX component library (accordions, callouts, cards, steps, tabs, code blocks, API-reference snippets, and more) for authoring docs pages.

    8.2k GitHub stars~2.8k tokensUpdated yesterday
    Auto-check passed

  • Fern Navigation

    ai-dynamo/dynamo

    Knowledge of Fern's site-level navigation and structure configuration — how a docs site is organized in docs.yml (and product/version .yml files) using sections, pages, folders, tabs, tab variants…

    8.2k GitHub stars~2.2k tokensUpdated yesterday
    Auto-check passed

  • Dynamo Frontend Benchmark

    ai-dynamo/dynamo

    Benchmark and profile the Dynamo frontend (dynamo.frontend HTTP + tokenizer + KV router) against mock workers (dynamo.mocker).

    8.2k GitHub stars~3.5k tokensUpdated yesterday
    Auto-check: notes

  • Configure Aiperf Benchmark

    ai-dynamo/dynamo

    Selects and freezes a question-driven AIPerf workload, objective, load policy, and Kubernetes execution manifest for a successfully deployed Dynamo candidate.

    8.2k GitHub stars~1.7k tokensUpdated yesterday
    Auto-check passed

  • Debug Session

    ai-dynamo/dynamo

    Sets up a structured debugging session for a Dynamo bug — pull the report from a Linear ticket, GitHub issue, or pasted text, capture the environment, create a persistent worklog markdown file, and…

    8.2k GitHub stars~1.2k tokensUpdated yesterday
    Auto-check passed

Works with

Categories

Questions about Dynamo Agent Harness

What does Dynamo Agent Harness do?

Drives persistent Claude Code, Codex, or OpenCode agent sessions through a Dynamo OpenAI/Anthropic-compatible endpoint over Agent Client Protocol (ACP). Dynamo Agent Harness is an agent skill from ai-dynamo/dynamo. Drives persistent Claude Code, Codex, or OpenCode agent sessions through a Dynamo OpenAI/Anthropic-compatible endpoint over Agent Client Protocol (ACP).

When should I use Dynamo Agent Harness?

Dynamo Agent Harness fits situations like: an agent must delegate a bounded task to another coding-agent harness running a model served by Dynamo; continue that harness across multiple turns; exercise tool calls; validate agent request traces.

How do I install Dynamo Agent Harness in Claude Code?

Run `npx skills add ai-dynamo/dynamo --skill dynamo-agent-harness -a claude-code`. Or copy the skill folder (.agents/skills/dynamo-agent-harness in ai-dynamo/dynamo) into .claude/skills/dynamo-agent-harness in your project. Claude Code loads it when a task matches its description.

How do I install Dynamo Agent Harness in Codex?

Run `npx skills add ai-dynamo/dynamo --skill dynamo-agent-harness -a codex`. Or copy the skill folder (.agents/skills/dynamo-agent-harness in ai-dynamo/dynamo) into .agents/skills/dynamo-agent-harness in your project. Codex loads it when a task matches its description.

Can I use Dynamo Agent Harness in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add ai-dynamo/dynamo --skill dynamo-agent-harness -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/dynamo-agent-harness, .gemini/skills/dynamo-agent-harness, .github/skills/dynamo-agent-harness and .opencode/skills/dynamo-agent-harness in your project.

What does Dynamo Agent Harness need to run?

Going by SKILL.md and its folder, Dynamo Agent Harness needs Python for the scripts in its folder, the command-line tools its instructions call (jq and opencode) and credentials named DYNAMO_API_KEY. Our summary lists: Python 3; Node.js; A credential in DYNAMO_API_KEY.

Does Dynamo Agent Harness access the network?

SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.

Is Dynamo Agent Harness safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. The check reads SKILL.md only: the scripts in the folder are not scanned, so read them before running anything.

What licence does Dynamo Agent Harness use?

Dynamo Agent Harness is published under the Apache-2.0 licence (declared in SKILL.md). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Dynamo Agent Harness use?

About 1.2k tokens (SKILL.md is roughly 4.6k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to Dynamo Agent Harness?

Skills that share tags, products or a category with Dynamo Agent Harness: Chroma Vector Database (Orchestra-Research/AI-Research-SKILLs, 13k stars), Codebase Management (giancarloerra/SocratiCode, 3.3k stars), CLIP Image-Text Matching (Orchestra-Research/AI-Research-SKILLs, 13k stars) and Docs Planner (strands-agents/harness-sdk, 8.7k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Dynamo Agent Harness?

ai-dynamo (a GitHub organization) maintains it in ai-dynamo/dynamo, which has 8,250 GitHub stars. The repository holds 27 skills in this directory. The repository was last updated on October 9, 2026.

Source: ai-dynamo/dynamo on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.