autoresearch

autoresearch is a command for coding agents from FlorianBruniaux/claude-code-plugins. It costs 24 tokens per session (2,434 once invoked), scanned A, original, MIT.

Autonomous improvement loop — scan codebase metrics, scaffold experiment files, run agent-driven iterations until metric improves.

Command

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 commands/florianbruniaux/claude-code-plugins/autoresearch
Clone the repo
git clone --depth 1 https://github.com/FlorianBruniaux/claude-code-plugins

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 autoresearch

README.md
[![agentmods](https://agentmods.dev/badge/commands/florianbruniaux/claude-code-plugins/autoresearch.svg)](https://agentmods.dev/commands/florianbruniaux/claude-code-plugins/autoresearch)
Your own site
<a href="https://agentmods.dev/commands/florianbruniaux/claude-code-plugins/autoresearch"><img src="https://agentmods.dev/badge/commands/florianbruniaux/claude-code-plugins/autoresearch.svg" alt="Measured on agentmods" height="20"></a>
Per session 24 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,434 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin unknown 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 $0.00024 $0.02434
Opus 5 $0.00012 $0.01217
Sonnet 5 $0.00005 $0.00487
Haiku 4.5 $0.00002 $0.00243

Measured today against content hash 36608317a0c7, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

autoresearch 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 today.

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/session-tools/commands/autoresearch.md · 294 lines

How it starts

The opening of the file, as written. The whole thing — 294 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Autoresearch — Autonomous Improvement Loop

Scan codebase quality metrics, propose improvement loops, and run autonomous agent iterations. Inspired by karpathy/autoresearch — adapted from ML research to code quality.

Concept: The agent proposes a code change, runs the measurement, keeps the change if the metric improved, reverts via git reset if not, and repeats until manually stopped.

Time: Scan ~30s | Per iteration: depends on scope | Loop: runs indefinitely until you stop it


Mode 1: Scan (default)

Measure current state, detect existing loops, propose next actions.

Instructions

Run the following metrics and display a prioritized proposal table.

Step 1: Measure codebase metrics

Adapt grep patterns to your project's conventions. These are TypeScript defaults — adjust for your stack.

# M1: Function declarations (prefer arrow functions)
M1=$(grep -r "export function " src/ --include="*.ts" --include="*.tsx" -l 2>/dev/null | wc -l | tr -d ' ')

# M2: Interface declarations (prefer type aliases)
M2=$(grep -r "export interface " src/ --include="*.ts" --include="*.tsx" -l 2>/dev/null | wc -l | tr -d ' ')

# M3: ESLint disables
M3=$(grep -r "eslint-disable" src/ --include="*.ts" --include="*.tsx" -l 2>/dev/null | wc -l | tr -d ' ')

# M4: Type casts to any
M4=$(grep -r " as any" src/ --include="*.ts" --include="*.tsx" -l 2>/dev/null | wc -l | tr -d ' ')

# M5: TODO comments
M5=$(grep -r "// TODO" src/ --include="*.ts" --include="*.tsx" -l 2>/dev/null | wc -l | tr -d ' ')

Step 2: Detect existing loops

for dir in scripts/autoresearch/loop-*/; do
  [ -d "$dir" ] || continue
  LOOP_NAME=$(basename "$dir")
  # Check if loop has results
  if [[ -f "$dir/results.tsv" ]]; then
    ITERS=$(wc -l < "$dir/results.tsv" | tr -d ' ')
    BEST=$(sort -t$'\t' -k2 -n "$dir/results.tsv" | head -1 | cut -f2)
    echo "ACTIVE:$LOOP_NAME:iterations=$ITERS:best=$BEST"
  else
    echo "SCAFFOLDED:$LOOP_NAME"
  fi
done

Read the full file on GitHub · 294 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. today First seen · 294 lines · 24 tokens per session scan A 36608317a0c7

Subscribe to this mod's changes

autoresearch is a command published in the GitHub repository FlorianBruniaux/claude-code-plugins (40 stars, last pushed 2d ago), licensed MIT. It adds 24 tokens to every session and 2,434 once invoked, about $0.0001 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 2026-09-04.