agent-evolutions AGENTS.md

Instructions for agent-evolutions, a workflow that creates and tests multiple versions of a change before applying the best one. It stores the work, results, and machine state in project files.

In plain words
What is it for?
Use it to capture an objective, generate variant implementations, compare them using exit codes and a numeric rubric, and apply the selected result.
Why use it?
It gives coding agents a shared process for running controlled experiments and keeping each candidate version separate.

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/fmind/agent-evolutions/agents-md
Clone the repo
git clone --depth 1 https://github.com/fmind/agent-evolutions

Made for: Codex, OpenCode.

Per session 1,319 This file is loaded in full into every session.
When invoked 1,319 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 $0.01319 $0.01319
Opus 5 $0.00660 $0.00660
Sonnet 5 $0.00264 $0.00264
Haiku 4.5 $0.00132 $0.00132

Measured yesterday against content hash 2975df656eb1, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

agent-evolutions 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 yesterday.

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 · 48 lines

How it starts

The opening of the file, as written. The whole thing — 48 lines — stays where its author put it; the contents beside it link to each section on GitHub.

AGENTS.md

Overview

  • agent-evolutions ships three slash-command skills — /new-agent-evolution, /run-agent-evolution, /apply-agent-evolution — that together drive a coding agent through a genetic exploration loop: gather a verifiable objective, generate variants in parallel batches, learn from each generation, pick the winner by exit codes and numeric rubric, port the winner into the repo. One skill per phase (Capture / Run / Apply); each refuses to operate outside its phase based on evolution.yaml field presence.
  • Skills are file-based and harness-agnostic: the same skills/ tree works in Claude Code, Gemini CLI, GitHub Copilot, and OpenCode.
  • End-user state lives in .agents/evolutions/<id>-<slug>/: evolution.yaml (machine state, validated against evolution.schema.json), EVOLUTION.md (single human/agent surface — TL;DR, Brief, Variants, Results; the §Brief section is loaded verbatim by every variant sub-agent), variants/v<n>/workspace/ (per-variant isolated checkout or file copy), variants/v<n>/result.json (sub-agent output, validated against result.schema.json).
  • State is derived from field presence, not an explicit step machine: empty variants[] = ready (run /new-agent-evolution first); has variants but no winner = running; has winner but no applied = evaluated; has applied = done. Each skill checks reality on entry and refuses if invoked in the wrong phase — no step enum, no pause flag.

Conventions

  • Line caps. Skill files load into agent context on every run; verbosity directly costs context budget. Targets:

    • skills/<name>/SKILL.md — ≤ 100 lines per file (each phase skill is self-contained).
    • Project-root markdown (README.md, AGENTS.md) — uncapped, but tighten when it sprawls.

    When the file approaches its cap, prefer cutting motivational/explanatory prose over removing structural rules. Each rule should be stated once, in the imperative, in the file the agent loads when it needs it. The "why" lives in commit messages and PR descriptions.

Read the full file on GitHub · 48 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. yesterday First seen · 48 lines · 1,319 tokens per session scan A 2975df656eb1

Subscribe to this mod's changes

agent-evolutions AGENTS.md is an instructions file published in the GitHub repository fmind/agent-evolutions (1 stars, last pushed 3mo ago), licensed MIT. It adds 1,319 tokens to every session, about $0.0066 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.