evolve-skill

evolve-skill is a skill for Claude Code, Codex from taneltaluri/evolve-skill. It costs 128 tokens per session (4,013 once invoked), scanned A, original, MIT.

A measurement-based process for improving SKILL.md files, which are instruction files for coding agents. It tests whether changes improve results without overfitting or changing the skill’s purpose.

In plain words
What is it for?
Use it to evaluate, compare, and carefully optimize a collection of skills, especially when you have many skills to maintain.
Why use it?
It helps distinguish real improvements from score fluctuations and prevents automated edits from making a skill worse over time.

Skill for Claude CodeCodex

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/taneltaluri/evolve-skill/evolve-skill
Any agent
npx skills add taneltaluri/evolve-skill --skill evolve-skill
Clone the repo
git clone --depth 1 https://github.com/taneltaluri/evolve-skill

Made for: Claude Code, Codex.

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 evolve-skill

README.md
[![agentmods](https://agentmods.dev/badge/skills/taneltaluri/evolve-skill/evolve-skill.svg)](https://agentmods.dev/skills/taneltaluri/evolve-skill/evolve-skill)
Your own site
<a href="https://agentmods.dev/skills/taneltaluri/evolve-skill/evolve-skill"><img src="https://agentmods.dev/badge/skills/taneltaluri/evolve-skill/evolve-skill.svg" alt="Measured on agentmods" height="20"></a>
Per session 128 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 4,013 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 $0.00128 $0.04013
Opus 5 $0.00064 $0.02006
Sonnet 5 $0.00026 $0.00803
Haiku 4.5 $0.00013 $0.00401

Measured 4d ago against content hash 8fdf46e2e1a7, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

evolve-skill 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.

SKILL.md · 410 lines

How it starts

The opening of the file, as written. The whole thing — 410 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Evolve Skill

A measurement-first skill optimizer. Inspired by Karpathy's autoresearch and the darwin-skill project, but rebuilt around one conviction: if you cannot trust your measurement, you cannot trust your optimization.

Every decision this skill makes is gated by whether the measurement is reliable enough to justify it.


What This Skill Is Not

This is not a formatting linter. This is not a prompt rewriter. This is not an autonomous agent that will rewrite your skills in the background.

This is a disciplined experiment loop with strict gates around scoring noise, overfitting, and function drift — the three failure modes that make most "auto-optimize" tools produce worse skills over time.

If you want a quick skill tidy-up, use a simpler tool. If you have 20+ skills and need to know which ones are actually improving, this is for you.


Core Philosophy — The Three Gates

Every optimization cycle passes through three gates. If any gate fails, the cycle stops or rolls back.

Gate 1 — Measurement Stability (can we trust the score?)

Before any optimization runs, score the baseline skill three independent times using three separate sub-agents. Compute the standard deviation. If SD > 2.0 points, the rubric is too loose and no optimization may proceed until the rubric is tightened.

A ratchet mechanism with unreliable measurement is not a ratchet — it is a random walk pretending to be progress.

Gate 2 — Effect Size (is this improvement real?)

A keep decision requires Δscore ≥ max(3, 2×SD) where SD is the measurement noise from Gate 1. A skill that "improved" by 1.5 points when noise is ±2 points did not improve.

Gate 3 — Function Preservation (is it still the same skill?)

Before and after every change, extract the skill's core functions as a bullet list (via a sub-agent). If function overlap drops below 80%, the change is rejected regardless of score. A higher-scoring skill that silently lost a capability is worse, not better.

Read the full file on GitHub · 410 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. 4d ago First seen · 410 lines · 128 tokens per session scan A 8fdf46e2e1a7

Subscribe to this mod's changes

evolve-skill is a skill published in the GitHub repository taneltaluri/evolve-skill (2 stars, last pushed 4mo ago), licensed MIT. It adds 128 tokens to every session and 4,013 once invoked, about $0.0006 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.

Related

Other skills, from other repositories

create-notes

Write a note to {shareddir}/notes/ that future agents can actually act on. Use after every coral eval, when a heartbeat (reflect / consolidate / pivot) asks for a note, or when you discover a grader / build / runtime issue that future agents will hit. Covers 4 note variants (experiment / infra / focus / synthesis)…

Human-Agent-Society/CORAL · 217 tokens

arbor-agent-executor

Executor-dispatch phase for Arbor. Use when implementing an Idea Tree node through RunExecutor or RunExecutorParallel semantics: isolated git worktree, executor prompt construction, eval metadata injection, RunTraining policy, smoke/full evaluation, report parsing, artifact persistence, tree update, and insight…

RUC-NLPIR/Arbor · 63 tokens

coral-new-task

End-to-end recipe for adding a new task under examples/ — the three pieces that have to line up (task.yaml, seed/, and grader/), what to put in each, the TaskGrader API surface, the coral validate → smoke-test loop, and the common mistakes (repopath pointing at the wrong dir, score direction backwards, hidden answer…

Human-Agent-Society/CORAL · 138 tokens

arbor-agent-ideate

Strict IDEATE-stage skill for Arbor. Use immediately after TreeView(format="constraints") when drafting Idea Tree nodes, enforcing the ideadrafting and firstprinciplesprobe behavior, depth-aware idea levels, four-line TreeAddNode hypotheses, conflict checks, and self-filtering against shallow tweaks.

RUC-NLPIR/Arbor · 66 tokens

arbor-agent-plugins-hitl-budget

Domain adaptation, human-in-the-loop, and budget policy phase for Arbor. Use when a run mentions plugins, plugin profiles, mlekaggle, evalcontract, protectedpaths, requiredoutputs, lifecycle hooks, convergence, budgetpolicy, RunTraining stages, AskUser, or interaction modes auto/direction/review/collaborative.

RUC-NLPIR/Arbor · 76 tokens

organize-files

Organize the shared notes directory when it becomes hard to navigate. Restructure within research/ and experiments/, deduplicate, update index.md.

Human-Agent-Society/CORAL · 32 tokens