autoresearch

A command that starts or resumes one recorded experiment intended to improve a chosen measurement in a code repository. Repeated unattended runs require an explicit loop request.

In plain words
What is it for?
Use it to set up or continue an experiment, pause it with an off marker, and run a single optimization iteration by default.
Why use it?
It makes optimization work repeatable by preserving the objective, previous results, and the code changes tried so far.

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/baleen37/bstack/autoresearch
Clone the repo
git clone --depth 1 https://github.com/baleen37/bstack
Per session 7 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 438 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 $0.00007 $0.00438
Opus 5 $0.00003 $0.00219
Sonnet 5 $0.00001 $0.00088
Haiku 4.5 $0.00001 $0.00044

Measured yesterday against content hash 53b95d6c3c10, 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 yesterday.

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/autoresearch/commands/autoresearch.md · 52 lines

What it actually says

Autoresearch Command

You are starting or resuming an autonomous experiment. The default is one iteration and then stop. Use loop only when the user explicitly requests repeated unattended iterations.

Handle arguments

Arguments: $ARGUMENTS

If arguments = "off"

Create a .autoresearch/off sentinel file in the current directory:

mkdir -p .autoresearch && touch .autoresearch/off

Then tell the user autoresearch mode is paused. It can be resumed by running /autoresearch again (which will delete the sentinel).

If .autoresearch/autoresearch.md exists in the current directory (resume)

This is a resume. Do the following:

  1. Delete .autoresearch/off if it exists
  2. Read .autoresearch/autoresearch.md to understand the objective, constraints, and what's been tried
  3. Read .autoresearch/results.jsonl to reconstruct state:
    • Count total runs, kept, discarded, crashed
    • Find the baseline metric (first data row)
    • Find the best metric and which commit achieved it
    • Note which secondary metric columns are being tracked
  4. Read recent git log: git log --oneline -20
  5. If .autoresearch/ideas.md exists, read it for experiment inspiration
  6. Perform exactly one next experiment, then stop. If the arguments explicitly include loop, use .autoresearch/loop.sh with the requested runtime and iteration limit.

If .autoresearch/autoresearch.md does NOT exist (fresh start)

  1. Delete .autoresearch/off if it exists
  2. Invoke the autoresearch skill to set up the experiment from scratch
  3. If arguments were provided (other than "off"), use them as the goal description to skip/answer the setup questions
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. yesterday First seen · 52 lines · 7 tokens per session scan A 53b95d6c3c10

Subscribe to this mod's changes

autoresearch is a command published in the GitHub repository baleen37/bstack (4 stars, last pushed 12d ago), licensed MIT. It adds 7 tokens to every session and 438 once invoked, about $0.0000 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.