calibrate-eval-infrastructure

calibrate-eval-infrastructure is a skill for Claude Code, Codex from Hoja-Solutions/agent-stdlib. It costs 147 tokens per session (828 once invoked), scanned A, original, MIT.

A guide for setting up the computers and containers used to measure coding agents. An agentic coding benchmark is a repeatable test of how well an agent solves software tasks, and infrastructure noise is variation caused by the test environment rather than the agent.

In plain words
What is it for?
Configuring benchmark containers, checking resource limits, measuring runtime variation and choosing headroom based on observed results when running SWE-bench or similar evaluations.
Why use it?
Incorrect memory or CPU limits can make a process fail or change benchmark scores, making model comparisons unreliable. It helps separate environment problems from actual agent performance.

Skill for Claude CodeCodex

Part of the agent-stdlib plugin — 14 skills, 2 commands, 1 agent, 2 hooks, 2 MCP servers 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/hoja-solutions/agent-stdlib/calibrate-eval-infrastructure
Any agent
npx skills add Hoja-Solutions/agent-stdlib --skill calibrate-eval-infrastructure
Clone the repo
git clone --depth 1 https://github.com/Hoja-Solutions/agent-stdlib

Made for: Claude Code, Codex.

Or install agent-stdlib, the plugin that ships this one along with the rest of its 14 skills, 2 commands, 1 agent, 2 hooks, 2 MCP servers.

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 calibrate-eval-infrastructure

README.md
[![agentmods](https://agentmods.dev/badge/skills/hoja-solutions/agent-stdlib/calibrate-eval-infrastructure.svg)](https://agentmods.dev/skills/hoja-solutions/agent-stdlib/calibrate-eval-infrastructure)
Your own site
<a href="https://agentmods.dev/skills/hoja-solutions/agent-stdlib/calibrate-eval-infrastructure"><img src="https://agentmods.dev/badge/skills/hoja-solutions/agent-stdlib/calibrate-eval-infrastructure.svg" alt="Measured on agentmods" height="20"></a>
Per session 147 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 828 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 $0.00147 $0.00828
Opus 5 $0.00073 $0.00414
Sonnet 5 $0.00029 $0.00166
Haiku 4.5 $0.00015 $0.00083

Measured 4d ago against content hash 88b944d7ac02, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

calibrate-eval-infrastructure 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 4d ago.

The scan reads SKILL.md. This mod also ships 1 executable file (scripts/noise_check.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.

skills/calibrate-eval-infrastructure/SKILL.md · 66 lines

How it starts

The opening of the file, as written. The whole thing — 66 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Calibrate eval infrastructure

Source: Quantifying infrastructure noise in agentic coding evals. This had no packaged skill anywhere; the topic existed only as the article and a few summaries.

Container configuration can move an agentic coding benchmark by 6 or more points. That is larger than the gap between the top models, which means a careless resource setting can rank a worse model above a better one. Treat resource configuration as an experimental variable you control and report, on the same footing as prompt format and sampling temperature.

The mistake that causes most of it

A container has two separate numbers, and pinning them together is the trap:

  • a guaranteed allocation, the floor the workload always has, and
  • a hard kill threshold, the ceiling past which the runtime kills the process.

Set the floor equal to the ceiling and the workload has zero headroom. A normal memory spike crosses the line and the process dies an OOM death that looks like the agent failing the task. It was not the agent. It was the box.

Give the two numbers separate values and leave a band between them.

Size the band empirically

Do not guess the headroom. Sweep it and measure whether the score still moves:

  1. Run the benchmark at several ceilings (a useful starting reference is roughly 3x the baseline ceiling; Anthropic's sweep cut infra errors from 5.8% to 2.1% with negligible score change at that point).
  2. Run each configuration several times, because one run per config cannot separate signal from noise.
  3. Test whether the score differences across configurations are distinguishable from noise. If they are not, you have found a band wide enough that the infrastructure no longer decides the outcome.

The bundled script runs that test:

python scripts/noise_check.py --scores scores.json

It reports a permutation p-value across your configs. A high p-value (the article treats roughly 0.40 and above as within noise) means resource configuration is no longer moving the score, so any model differences you measure are real. A low p-value means the infrastructure is still talking, and you need a wider band before you trust a single number. See scripts/noise_check.py for the input format.

Read the full file on GitHub · 66 lines

Files

What ships with it

1 file 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. 4d ago First seen · 66 lines · 147 tokens per session scan A 88b944d7ac02

Subscribe to this mod's changes

calibrate-eval-infrastructure is a skill published in the GitHub repository Hoja-Solutions/agent-stdlib (1 stars, last pushed 1mo ago), licensed MIT. It adds 147 tokens to every session and 828 once invoked, about $0.0007 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.