compare-harnesses

compare-harnesses is a skill for Claude Code, Codex from ruvnet/metaharness. It costs 66 tokens per session (1,173 once invoked), scanned A, original, MIT.

A command that compares two scaffolded agent harnesses and reports differences in their metadata, supported host adapters, and files. A harness is the generated project that lets an agent run in a particular environment.

In plain words
What is it for?
Comparing a fork with an upstream template, investigating why two generated harnesses differ, and checking byte-level file fingerprints in automated tests.
Why use it?
It quickly shows whether two harnesses are identical or where they have drifted apart, including changes to individual files.

Skill for Claude CodeCodex

Written for Claude Code and Codex: a Claude Code plugin manifest, but also reads ~/.codex or $CODEX_HOME. Also seen: mentions Codex; built for openclaw.

Part of the agent-harness-generator plugin — 14 skills, 1 plugin shipped together

Good fit Comparing a fork with an upstream template, investigating why two generated harnesses differ, and checking byte-level file fingerprints in automated tests.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/ruvnet/metaharness/compare-harnesses
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.

Any agent
npx skills add ruvnet/metaharness --skill compare-harnesses
Clone the repo
git clone --depth 1 https://github.com/ruvnet/metaharness

Made for: Claude Code, Codex.

Or install agent-harness-generator, the plugin that ships this one along with the rest of its 14 skills, 1 plugin.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for compare-harnesses

README.md
[![agentmods](https://agentmods.dev/badge/skills/ruvnet/metaharness/compare-harnesses.svg)](https://agentmods.dev/skills/ruvnet/metaharness/compare-harnesses)
Your own site
<a href="https://agentmods.dev/skills/ruvnet/metaharness/compare-harnesses"><img src="https://agentmods.dev/badge/skills/ruvnet/metaharness/compare-harnesses.svg" alt="Measured on agentmods" height="20"></a>
Per session 66 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,173 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe. Third-party audits
  • NVIDIA SkillSpector pass 7 Sept 2026
How audits are shown
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 5.1 $0.00066 $0.01173
Opus 5 $0.00033 $0.00587
Sonnet 5 $0.00013 $0.00235
Haiku 4.5 $0.00007 $0.00117

Measured 8d ago against content hash 4a21e5bb735e, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-07, from the pricing page.

Security

Grade A, and why

compare-harnesses 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 8d 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.

.claude-plugin/skills/compare-harnesses/SKILL.md · 116 lines

How it starts

The opening of the file, as written. The whole thing — 116 lines — stays where its author put it; the contents beside it link to each section on GitHub.

compare-harnesses

Codex skill: diff two scaffolded harnesses — the ADR-031 Bundle JSON Pattern surfaced through Codex (iter 105 → iter 109).

What it does

Two-harness diff. Useful when you've forked an upstream template, when a support ticket says "mine and theirs scaffolded different things", or when a CI script wants a byte-equality check between a candidate scaffold and a known-good baseline.

Reports three sections:

  1. Manifest meta — same name? same kernel? same surface?
  2. Hosts — which host adapters each side ships (claude-code / codex / pi-dev / hermes / openclaw / rvm)
  3. Files — added / removed / changed (per-file SHA-256 fingerprints; the cheapest possible byte-equality test)
Verdict Exit Meaning
IDENTICAL 0 meta + hosts + every file fingerprint matches
DRIFT 1 at least one of meta / hosts / files differs
no-manifest-in-a 2 a/.harness/manifest.json missing
no-manifest-in-b 2 b/.harness/manifest.json missing
no-manifest-in-either 2 both sides missing manifest

Usage from Codex

/compare-harnesses a=./my-fork b=./upstream
/compare-harnesses a=./my-fork b=./upstream bundle=true

Equivalent CLI

harness compare ./my-fork ./upstream                  # text output
harness compare ./my-fork ./upstream --bundle         # ADR-031 schema-1 JSON

The --bundle form (ADR-031) emits a schema-1 JSON envelope so CI scripts can json-parse the verdict without re-parsing human text. The envelope shape is shared with harness diag --bundle, harness export-config, and harness audit --bundle:

{
  "schema": 1,
  "generatedAt": "2026-06-15T...",
  "a": "/path/to/a",
  "b": "/path/to/b",
  "meta": { "sameKernel": true, "sameSurface": true, "sameName": false },
  "hosts": { "a": ["claude-code"], "b": ["claude-code", "codex"], "verdict": "PASS" | "FAIL" },
  "files": { "added": [...], "removed": [...], "changed": [...] },
  "identical": false,
  "exitCode": 1
}

Read the full file on GitHub · 116 lines

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. 8d ago First seen · 116 lines · 66 tokens per session scan A 4a21e5bb735e

Subscribe to this mod's changes

compare-harnesses is a skill published in the GitHub repository ruvnet/metaharness (633 stars, last pushed today), licensed MIT. It adds 66 tokens to every session and 1,173 once invoked, about $0.0003 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 2026-08-30.