Autoresearch Agent
Plugin Verified ActiveAutonomous experiment loop that optimizes any file by a measurable metric. 5 slash commands, 8 evaluators, configurable loop intervals (10min to monthly).
To enable users to systematically optimize files or content for specific metrics through an automated, iterative experimentation process.
Features
- Autonomous experiment loop for file optimization
- Configurable loop intervals (10min to monthly)
- Supports various optimization targets (code speed, size, quality, content)
- Interactive setup and status dashboard
- Git integration for tracking changes and reverts
Use Cases
- Optimizing code performance or size
- Improving content quality like headlines or copy
- Running automated experiments overnight or over weeks
- Systematically refining prompts for AI agents
Non-Goals
- Performing remote code execution or deployments
- Managing external dependencies or secrets
- Operating on files outside the specified target or experiment directory
- Replacing manual code reviews or architectural design
Installation
First, add the marketplace
/plugin marketplace add alirezarezvani/claude-skills/plugin install autoresearch-agent@claude-code-skillsContains 4 extensions
Skill (4)
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 command that outputs a metric, and a git repo.
Start an autonomous experiment loop with user-selected interval (10min, 1h, daily, weekly, monthly). Uses CronCreate for scheduling.
Resume a paused experiment. Checkout the experiment branch, read results history, continue iterating.
Set up a new autoresearch experiment interactively. Collects domain, target file, eval command, metric, direction, and evaluator.
Quality Score
VerifiedTrust Signals
Similar Extensions
Karpathy Coder
100Active coding discipline enforcer based on Karpathy's 4 principles: surface assumptions, keep it simple, make surgical changes, define verifiable goals. Ships 4 Python tools (complexity_checker, diff_surgeon, assumption_linter, goal_verifier), a review agent, /karpathy-check slash command, and a pre-commit hook. All tools stdlib-only.
Train Sentence Transformers
99Train or fine-tune sentence-transformers models across all three architectures: SentenceTransformer (bi-encoder embeddings), CrossEncoder (rerankers), and SparseEncoder (SPLADE). Covers loss selection, hard-negative mining, evaluators, distillation, LoRA, Matryoshka, and Hugging Face Hub publishing.