autoevolve-optimizer

An automated loop for improving selected configuration files. It changes a configuration, scores it against fixed test cases, and keeps a change only when it improves the score.

In plain words
What is it for?
Tuning the detection-index or context-router configuration after enough real outcomes have been collected. It is also for running repeatable, no-LLM-cost comparisons of configuration changes.
Why use it?
It prevents configuration experiments from making things worse. Built-in checks stop changes when there is not enough real usage data and restore any result below the previous baseline.

Agent

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/primeline-ai/evolving-lite/autoevolve-optimizer
Clone the repo
git clone --depth 1 https://github.com/primeline-ai/evolving-lite
Per session 36 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,185 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 $0.00036 $0.01185
Opus 5 $0.00018 $0.00593
Sonnet 5 $0.00007 $0.00237
Haiku 4.5 $0.00004 $0.00119

Measured 2d ago against content hash fe9350500a26, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

autoevolve-optimizer 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 2d 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/autoevolve-optimizer.md · 96 lines

How it starts

The opening of the file, as written. The whole thing — 96 lines — stays where its author put it; the contents beside it link to each section on GitHub.

AutoEvolve Optimizer

You iteratively improve a target config file by proposing mutations, scoring them against fixed test cases (zero LLM cost), and keeping only improvements. Two gates ship as CODE and are NOT yours to skip:

  • mutation-eligibility gate (before you start): refuses to run unless the global switch is on, the target is enabled, and enough real outcomes have accumulated. A fresh install with no usage data has nothing to tune yet.
  • deterministic persist-gate (after each score): re-scores the live config against a pre-mutation snapshot and auto-reverts any below-baseline result, independent of your own revert. A regression cannot stick even if you forget.

Paths are under ${CLAUDE_PLUGIN_ROOT}. The scorer is scripts/autoevolve-scorer.py; helpers are scripts/v2_runner_helpers.py.

Step 0 - Eligibility (MANDATORY before any mutation)

python3 scripts/autoevolve-scorer.py mutation-gate {target}

Exit 0 = eligible, proceed. Exit 1 = blocked (global off, target disabled, or fewer than the MVP sample threshold of real outcomes). If blocked, STOP and report the reason; do not mutate anything.

Core Loop

Read _autoevolve/config.json -> confirm {target} is enabled + read its safety block
Create a feature branch: autoevolve/{target}/{YYYY-MM-DD-HHMMSS}   (NEVER main)
Run the scorer once to establish the baseline.

FOR each iteration (1 .. budget):
  1. READ the target file + test cases + last scorer failures
  2. SNAPSHOT before mutating:
       cp {target_file} _autoevolve/snapshots/pre-{target}-{ts}.json
  3. PROPOSE one specific mutation (Rule 1: exactly one change)
  4. APPLY via Edit
  5. SCORE: python3 scripts/autoevolve-scorer.py score {target}
  6. PERSIST-GATE (code-enforced revert backstop):
       python3 scripts/autoevolve-scorer.py persist-gate {target} \
         --snapshot _autoevolve/snapshots/pre-{target}-{ts}.json \
         --run-id {branch} --desc "{one-line mutation summary}"
     exit 0 = kept, exit 2 = auto-reverted (regression caught), exit 3 = skip
     (non-deterministic target). exit 4 = ERROR (scoring/restore failed - the
     gate did NOT run): STOP the loop and investigate, do not continue mutating.
  7. IF improved (gate kept + score up): git commit on the branch; log "+{delta}"
     IF not improved: ensure the file is restored (the gate does it on regression;
       you restore on a plateau/no-op). Record the rejected mutation:
         python3 scripts/v2_runner_helpers.py reject --target {target} \
           --run-id {branch} --description "{summary}" \
           --score-before {baseline} --score-after {new} --reason {regression|plateau}
  8. CHECK plateau: python3 scripts/autoevolve-scorer.py plateau {target}
     IF plateau AND >10 iterations used: STOP early.

Read the full file on GitHub · 96 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. 2d ago First seen · 96 lines · 36 tokens per session scan A fe9350500a26

Subscribe to this mod's changes

autoevolve-optimizer is an agent published in the GitHub repository primeline-ai/evolving-lite (48 stars, last pushed 15d ago), licensed MIT. It adds 36 tokens to every session and 1,185 once invoked, about $0.0002 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.

Related

Other agents, from other repositories

tdd-enforcer

Test-Driven Development enforcer. Strict RED-GREEN-REFACTOR discipline. ACTIVATE when user mentions: "tdd", "test first", "write tests", "failing test", "red green refactor", "test driven", "test before implementation". Works with Laravel (PHPUnit/Pest), Flutter (fluttertest), Vue (Vitest/Jest). Enforces test-first…

anilcancakir/claude-code-plugins · 92 tokens

idea-validator

Business idea validation specialist using proven scoring frameworks. Use for systematic idea evaluation, problem-solution fit assessment, SWOT analysis, and go/no-go recommendations. Invoke when users want to validate a business concept, assess founder-market fit, or get objective scoring on startup ideas.

anilcancakir/claude-code-plugins · 56 tokens

code-reviewer

Expert code reviewer for security vulnerabilities, bugs, logic errors, and code quality. Use PROACTIVELY when reviewing uncommitted changes, analyzing diffs, or checking for OWASP Top 10 issues. Ideal for pre-commit reviews in Laravel, Flutter, Vue, and TypeScript projects.

anilcancakir/claude-code-plugins · 64 tokens

doc-updater

Documentation specialist for updating inline docs (PHPDoc, JSDoc, TSDoc, VueDoc, DartDoc) and project documentation. Use when code changes need documentation updates. Fast and efficient for doc tasks.

anilcancakir/claude-code-plugins · 47 tokens

market-researcher

Deep market analysis specialist for TAM/SAM/SOM calculation, industry dynamics, and competitive intelligence. Use for comprehensive market sizing, Porter's Five Forces analysis, and identifying growth drivers. Invoke when users need market opportunity assessment, industry structure analysis, or detailed competitive…

anilcancakir/claude-code-plugins · 59 tokens

trend-analyzer

Market trend analysis specialist for identifying emerging patterns, timing assessment, and signal detection. Use for trend identification, Google Trends analysis, market timing evaluation, and opportunity window assessment. Invoke when users need to understand if timing is right, identify emerging opportunities, or…

anilcancakir/claude-code-plugins · 58 tokens