self-evolve CLAUDE.md

self-evolve CLAUDE.md is an instructions file for coding agents from Stealthy-McStealth/self-evolve. It costs 282 tokens per session, scanned A, original, MIT.

Instructions that make an assistant regularly check that it is following its loaded skills during long, multi-step tasks.

In plain words
What is it for?
Use them to announce relevant skills, check compliance at major steps, handle fallback instructions explicitly, and record useful lessons when creating new skills.
Why use it?
They reduce the chance that the assistant gradually ignores required procedures or improvises when a task becomes complicated.

Instructions file

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/stealthy-mcstealth/self-evolve/claude-md
Clone the repo
git clone --depth 1 https://github.com/Stealthy-McStealth/self-evolve

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 self-evolve CLAUDE.md

README.md
[![agentmods](https://agentmods.dev/badge/instructions/stealthy-mcstealth/self-evolve/claude-md.svg)](https://agentmods.dev/instructions/stealthy-mcstealth/self-evolve/claude-md)
Your own site
<a href="https://agentmods.dev/instructions/stealthy-mcstealth/self-evolve/claude-md"><img src="https://agentmods.dev/badge/instructions/stealthy-mcstealth/self-evolve/claude-md.svg" alt="Measured on agentmods" height="20"></a>
Per session 282 This file is loaded in full into every session.
When invoked 282 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 $0.00282 $0.00282
Opus 5 $0.00141 $0.00141
Sonnet 5 $0.00056 $0.00056
Haiku 4.5 $0.00028 $0.00028

Measured 5d ago against content hash 763f75b397f2, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

self-evolve CLAUDE.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 5d 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.md · 40 lines

What it actually says

harness adherence protocol

these instructions are always active. they're not optional and they don't require a slash command to engage. they exist because research shows that even when skills are loaded, models progressively drift from following them across long trajectories.

source: "harness updating is not harness benefit" (lin et al., arXiv:2605.30621)


before starting any multi-step task

state out loud (briefly) which loaded skills are relevant to this task. if no skills are relevant, say so. this makes it checkable.

at each major sub-task transition

ask: am i still following the loaded skill's procedure? if the answer is "i'm not sure" — re-read the skill before continuing.

if a skill has fallback steps

acknowledge them before starting, not after step 1 fails. "if X fails, i will try Y" — commit to this upfront.

if you notice yourself improvising around a skill

stop. re-read the skill. /reload-context if needed. improvising around loaded instructions is the failure mode. not forgetting.


on writing new skills

when a learning emerges from a conversation, encode it. /improve-urself exists for this. use it.

an insight that lives only in this conversation window is gone after the session. a skill persists.

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 · 40 lines · 282 tokens per session scan A 763f75b397f2

Subscribe to this mod's changes

self-evolve CLAUDE.md is an instructions file published in the GitHub repository Stealthy-McStealth/self-evolve (5 stars, last pushed 3mo ago), licensed MIT. It adds 282 tokens to every session, about $0.0014 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.