sigma AGENTS.md

sigma AGENTS.md is an instructions file for Codex, OpenCode from hututuQQQ/sigma. It costs 312 tokens per session, scanned A, original, MIT.

An AGENTS.md instruction file for the sigma project that sets rules for fair software benchmarks. It forbids changing behavior specifically to recognize or optimize for benchmark tasks or hidden test results.

In plain words
What is it for?
Use it to guide agents and test harnesses when running benchmark tasks, especially around task selection, prompts, retries, verifier feedback, and benchmark-related files.
Why use it?
It prevents benchmark-specific shortcuts from making results misleading and keeps the agent solving the stated programming problem generally.

Instructions file for CodexOpenCode

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 instructions/hututuqqq/sigma/agents-md
Clone the repo
git clone --depth 1 https://github.com/hututuQQQ/sigma

Made for: Codex, OpenCode.

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 sigma AGENTS.md

README.md
[![agentmods](https://agentmods.dev/badge/instructions/hututuqqq/sigma/agents-md.svg)](https://agentmods.dev/instructions/hututuqqq/sigma/agents-md)
Your own site
<a href="https://agentmods.dev/instructions/hututuqqq/sigma/agents-md"><img src="https://agentmods.dev/badge/instructions/hututuqqq/sigma/agents-md.svg" alt="Measured on agentmods" height="20"></a>
Per session 312 This file is loaded in full into every session.
When invoked 312 The same file — it is already loaded in full.
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 $0.00312 $0.00312
Opus 5 $0.00156 $0.00156
Sonnet 5 $0.00062 $0.00062
Haiku 4.5 $0.00031 $0.00031

Measured 4d ago against content hash 9eaf9a9d8606, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

sigma AGENTS.md 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.

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.

AGENTS.md · 24 lines

What it actually says

Agent Instructions

Benchmark Fairness

Benchmark-directed programming is prohibited. Treat it as a serious integrity violation, not as an optimization.

Agents and harness code must not:

  • Detect or branch on benchmark names, task IDs, task names, dataset names, verifier test names, package names, fixture names, artifact names, or known benchmark outputs.
  • Add task-specific smoke tests, prechecks, prompts, retries, cleanup rules, or service behavior based on benchmark/task identity.
  • Pass benchmark task identity, task hints, verifier failures, verifier traces, verifier stdout/stderr, rewards, scores, or hidden test details into the solving agent.
  • Retry a solving attempt using verifier feedback or any other post-verifier information.
  • Tune prompts, commands, package installs, assertions, file names, or heuristics to improve a known benchmark score instead of solving the user's stated task generally.
  • Keep "dev-only" benchmark shortcuts in the default agent, harness, CLI, or portable runtime paths.

Allowed benchmark code is limited to neutral infrastructure:

  • Selecting which benchmark task to run from an external runner.
  • Packaging and launching the agent without exposing benchmark identity to the agent.
  • Collecting logs, traces, summaries, verifier results, scores, and reports after the run for human inspection.
  • Generic validation that is derived from user-provided commands, summary-declared commands, changed-file syntax checks, or broadly applicable project conventions.

When in doubt, remove benchmark-specific behavior. Prefer a lower score with a fair, reusable harness over any score gained from benchmark knowledge.

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 · 24 lines · 312 tokens per session scan A 9eaf9a9d8606

Subscribe to this mod's changes

sigma AGENTS.md is an instructions file published in the GitHub repository hututuQQQ/sigma (11 stars, last pushed 15d ago), licensed MIT. It adds 312 tokens to every session, about $0.0016 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.