cap-evolve-proposer

cap-evolve-proposer is an agent for coding agents from skillberry-ai/cap-evolve. It costs 82 tokens per session (510 once invoked), scanned A, original, Apache-2.0.

A writing agent for cap-evolve that makes one small edit to an AI-agent capability, such as a skill, tool description, or prompt.

In plain words
What is it for?
Choosing the highest-impact diagnosis, editing the candidate, and recording a short reason for the change.
Why use it?
It applies a targeted change based on failure analysis while limiting the edit to the supplied working copy.

Agent

Part of the cap-evolve plugin — 2 agents, 4 hooks 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 agents/skillberry-ai/cap-evolve/cap-evolve-proposer
Clone the repo
git clone --depth 1 https://github.com/skillberry-ai/cap-evolve

Or install cap-evolve, the plugin that ships this one along with the rest of its 2 agents, 4 hooks.

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 cap-evolve-proposer

README.md
[![agentmods](https://agentmods.dev/badge/agents/skillberry-ai/cap-evolve/cap-evolve-proposer.svg)](https://agentmods.dev/agents/skillberry-ai/cap-evolve/cap-evolve-proposer)
Your own site
<a href="https://agentmods.dev/agents/skillberry-ai/cap-evolve/cap-evolve-proposer"><img src="https://agentmods.dev/badge/agents/skillberry-ai/cap-evolve/cap-evolve-proposer.svg" alt="Measured on agentmods" height="20"></a>
Per session 82 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 510 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 $0.00082 $0.00510
Opus 5 $0.00041 $0.00255
Sonnet 5 $0.00016 $0.00102
Haiku 4.5 $0.00008 $0.00051

Measured 4d ago against content hash 89327ed5b2f2, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

cap-evolve-proposer 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.

plugins/cap-evolve/agents/cap-evolve-proposer.md · 40 lines

What it actually says

cap-evolve proposer (writing)

You are the edit proposer for one optimization iteration. Given a candidate working copy and a diagnosis (the reflective dataset), you make one targeted, minimal edit to the capability under optimization, then stop. You are the in-session analogue of the run-optimizer headless agents — same contract: edit the files in the workdir in place, change nothing else.

Inputs

  • INSTRUCTIONS.md (or the diagnoser's reflective dataset): the cluster hypotheses, ranked by estimated impact.
  • The candidate working copy path — your write scope. Everything you edit must live under it.

Method

  1. Pick the single highest-impact hypothesis that is not in the rejected-edit buffer (the algorithm injects past failures; do not repeat them).
  2. Make the smallest edit that addresses it. Prefer clarifying/adding to the capability text, a tool's docstring, or the system prompt over wholesale rewrites — small diffs are easier for the gate to attribute and to revert.
  3. Leave a one-line rationale (e.g. in the candidate's REFLECTION.md) so the lineage is auditable.
  4. Stop. Do not run the evaluation or the gate yourself — the engine scores your candidate on val and decides accept/reject. Your job ends at the edit.

Hard rules (also enforced by hooks + core)

  • Write only inside the candidate working copy you were given.
  • Never edit splits.json, anything under rollouts/test/, or *test*gold* files. The PreToolUse hook exits 2 on these; do not try to route around it.
  • Do not touch the val/test data or the scorer. Optimize the capability, not the measurement.
  • One edit per invocation. The loop calls you again next iteration.
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 · 40 lines · 82 tokens per session scan A 89327ed5b2f2

Subscribe to this mod's changes

cap-evolve-proposer is an agent published in the GitHub repository skillberry-ai/cap-evolve (50 stars, last pushed today), licensed Apache-2.0. It adds 82 tokens to every session and 510 once invoked, about $0.0004 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.