Name: autoresearch-agent
Rating: 74.4 (1 reviews)
Author: alirezarezvani

autoresearch-agent

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.

Quality

92%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

The canonical home for this skill is autoresearch-agent in alirezarezvani/claude-skills

Quality

Content

85%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

The body is well-structured and highly actionable with a clear, validated experiment-loop workflow, but it is somewhat padded by redundant trigger/rules duplication and leans on references to scripts that are not actually bundled.

Suggestions

Ship the referenced bundle (scripts/setup_experiment.py, run_experiment.py, log_results.py, convert.sh) or remove the inline command references so the skill does not point at missing files.

Deduplicate 'Proactive Triggers' against 'Rules' — keep crash/evaluator-modification guidance in one place to trim tokens.

Move the evaluator tables and dashboard/export format details into a single references/ file referenced one level deep, leaving the body as an overview.

Dimension	Reasoning	Score
Conciseness	Largely lean and assumes Claude's competence (commands, configs, rules without basic-concept padding), but the 'Proactive Triggers' section restates rules already in 'Rules' (e.g., '5 consecutive crashes', 'modifying evaluate.py') and the dashboard/export blocks are illustrative — minor trims available.	4 / 5
Actionability	Provides copy-paste-ready, executable commands throughout (setup with full flags, 'git add ... && git commit', 'python scripts/run_experiment.py --experiment ... --single', an executable custom-evaluator Python example) covering the common cases.	5 / 5
Workflow Clarity	The 'Each Iteration' loop is a clearly sequenced 1-7 flow with explicit validation feedback (KEEP/DISCARD/CRASH), crash-handling and '5 consecutive crashes -> pause' checkpoints, plus Strategy Escalation and Self-Improvement feedback loops over runs.	5 / 5
Progressive Disclosure	Headers and sectioning are good, but the body references scripts and slash commands (scripts/setup_experiment.py, run_experiment.py, log_results.py, convert.sh, /ar:*) that do not exist as bundle files, and substantial content (evaluator tables, dashboard/export formats) is inlined rather than split into one-level-deep referenced files.	3 / 5
	Total	17 / 20 Passed

Description

100%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

The description is concise yet comprehensive: it states the concrete mechanic, lists natural trigger phrases covering multiple domains, and spells out prerequisites. It uses third person throughout and avoids vague fluff.

Dimension	Reasoning	Score
Specificity	Names multiple concrete actions ('edits a target file, runs a fixed evaluation, keeps improvements (git commit), discards failures (git reset), and loops indefinitely') with comprehensive coverage, matching the anchor-5 example.	5 / 5
Completeness	Explicitly answers both 'what' (the autonomous edit-evaluate-keep-discard-loop mechanic) and 'when' ('Use when: user wants to optimize...') with concrete trigger phrases, matching the anchor-5 example.	5 / 5
Trigger Term Quality	Comprehensive natural trigger phrases users would actually say ('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'), including synonyms.	5 / 5
Distinctiveness Conflict Risk	Occupies a clear niche (measurable-optimization loop with git commit/reset semantics); triggers are specific enough that it is unlikely to fire for unrelated skills, giving minimal conflict risk.	5 / 5
	Total	20 / 20 Passed

Validation

93%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 15 / 16 Passed

Validation for skill structure

Criteria	Description	Result
referenced_paths_exist	Referenced path issues: 11 missing	Warning

	Total	15 / 16 Passed

Repository: alirezarezvani/claude-skills
Path: .gemini/skills/autoresearch-agent/SKILL.md
Commit: 19392f7

Reviewed: ~3 hours ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.