autoresearch-agent

Elevated · Review

Autonomous experiment loop that optimizes any file by a measurable metric. Inspired by Karpathy's autoresearch. The agent edits a target file, runs a fixed evaluation, keeps improvements (git commit), discards failures (git reset), and loops indefinitely. Use when: user wants to optimize code speed, reduce bundle/image size, improve test pass rate, optimize prompts, improve content quality (headlines, copy, CTR), or run any measurable improvement loop. Requires: a target file, an evaluation c...

Governance Receipt

Signer
sovereign-claw-ed25519
Signed At
6/4/2026
Risk Tier
T2
Receipt Hash
5ddf541e
Manifest Hash
beef4d9e5553b6d95eea88378664f1d778a34c385368c46223ba49b69de7a5ab
Signature
3FCVlO6G
Root Public Key
349b0348

Skill Details

Gate Verdict
Elevated · Review
Publication State
published
Risk Tier
T2
Manifest Hash
beef4d9e

Install autoresearch-agent

Each bundle is served from the attested copy of autoresearch-agent, so the file you download is the one covered by the governance receipt on this page rather than whatever currently sits at the upstream source.

Claude Code skill

curl -L "https://atestiv.com/api/download/autoresearch-agent?format=claude-code" -o autoresearch-agent.skill.md

MCP tool definitions

curl -L "https://atestiv.com/api/download/autoresearch-agent?format=mcp" -o autoresearch-agent.mcp-tools.json

Raw attested bundle

curl -L "https://atestiv.com/api/download/autoresearch-agent?format=raw" -o autoresearch-agent.bundle.json

Verify the bundle against its manifest hash

shasum -a 256 autoresearch-agent.bundle.json

Provenance

  • Catalog identifier: autoresearch-agent
  • Harvested from the upstream repository.
  • The manifest hash below is what the signing key committed to. A bundle whose digest does not match it is not the attested autoresearch-agent.
  • Anything published here has been through the same gate, so a missing receipt is itself a signal rather than an omission.

More Skills