os-eval-runner

os-eval-runner is a skill for Claude Code, Codex from richfrem/agent-plugins-skills. It costs 64 tokens per session (942 once invoked), scanned A, original, MIT.

A tool for testing and scoring changes to an agent skill with Python evaluation scripts. It runs repeated improvement attempts and accepts or rejects them based on the results.

In plain words
What is it for?
Use it to evaluate, optimise, or run an autoresearch loop on a skill, with experiment files containing instructions, tests, and result records.
Why use it?
It gives skill changes a measurable test process, helping prevent unverified improvements or regressions.

Skill for Claude CodeCodex

Part of the agent-agentic-os plugin — 20 skills, 4 commands, 2 agents, 3 hooks shipped together

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/richfrem/agent-plugins-skills/os-eval-runner
Any agent
npx skills add richfrem/agent-plugins-skills --skill os-eval-runner
Clone the repo
git clone --depth 1 https://github.com/richfrem/agent-plugins-skills

Made for: Claude Code, Codex.

Or install agent-agentic-os, the plugin that ships this one along with the rest of its 20 skills, 4 commands, 2 agents, 3 hooks.

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 os-eval-runner

README.md
[![agentmods](https://agentmods.dev/badge/skills/richfrem/agent-plugins-skills/os-eval-runner.svg)](https://agentmods.dev/skills/richfrem/agent-plugins-skills/os-eval-runner)
Your own site
<a href="https://agentmods.dev/skills/richfrem/agent-plugins-skills/os-eval-runner"><img src="https://agentmods.dev/badge/skills/richfrem/agent-plugins-skills/os-eval-runner.svg" alt="Measured on agentmods" height="20"></a>
Per session 64 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 942 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 $0.00064 $0.00942
Opus 5 $0.00032 $0.00471
Sonnet 5 $0.00013 $0.00188
Haiku 4.5 $0.00006 $0.00094

Measured 5d ago against content hash be78ba4c14e0, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

os-eval-runner 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 5d ago.

The scan reads SKILL.md. This mod also ships 5 executable files (scripts/eval_runner.py, scripts/evaluate.py, scripts/generate_milestone.py, …), listed below but not scanned — reading those needs a real analyzer, not pattern matching.

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-agentic-os/skills/os-eval-runner/SKILL.md · 73 lines

How it starts

The opening of the file, as written. The whole thing — 73 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Skill Improvement Evaluator

Stateless evaluation engine that scores and gates skill improvement iterations using headless Python evaluation scripts.


Ownership Boundary (Critical)

What os-eval-runner owns (permanent, version-controlled with this skill)

  • Scoring scripts: ./scripts/evaluate.py, ./scripts/eval_runner.py
  • Scaffold script: ./scripts/init_autoresearch.py
  • Templates: ./assets/templates/autoresearch/ (program, evals, results, proposer prompt)

What lives with the target (deployed per experiment)

All experiment state deploys alongside the target (e.g. <experiment-dir>/references/program.md, <experiment-dir>/evals/evals.json, <experiment-dir>/evals/results.tsv). You MUST read the spec from <experiment-dir>/references/program.md and NOT fall back to engine-local config templates.


Phase 0: Intake Interview

Run this interview before starting any loop or evaluation. If enough information is provided in the initial prompt, skip the redundant questions.

  1. Q1 — What target skill are you evaluating? (Provide path to skill folder)
  2. Q2 — Where should the experiment files live? (Defaults to target skill directory)
  3. Q2b — What metric are you optimizing? (quality_score, f1, precision, recall, or heuristic)
  4. Q3 — What mode? (Loop mode for autonomous improvement vs QA mode for single diff validation)
  5. Q4 — (Loop mode) How many iterations? (Default: NEVER STOP)
  6. Q5 — Does evals.json exist? (If missing, scaffold from template)
  7. Q6 — Does program.md exist? (If missing, scaffold from template)
  8. Q7 — Does a baseline score exist? (If missing, run evaluate.py with --baseline)

Two Modes: Summarized

  • Mode 1: Autoresearch Loop: Autonomous iterative improvement. The agent identifies failure types, requests mutations via external proposer CLI (Copilot/Gemini), and runs the eval gate iteratively until the budget or target score is met.
  • Mode 2: Single-shot QA: Simple gate validation. Evaluates one specific proposed diff against the baseline and decides KEEP (exit 0) or DISCARD (revert, exit 1).

Read the full file on GitHub · 73 lines

Files

What ships with it

37 files 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. 5d ago First seen · 73 lines · 64 tokens per session scan A be78ba4c14e0

Subscribe to this mod's changes

os-eval-runner is a skill published in the GitHub repository richfrem/agent-plugins-skills (6 stars, last pushed yesterday), licensed MIT. It adds 64 tokens to every session and 942 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 2026-08-31.