lab:autoresearch
lab:autoresearch runs a self-improving loop that mutates plugin skills one change at a time, scores each against deterministic criteria, and commits improvements while reverting failures. Target the weakest skill or run a full sweep overnight with /loop.
lab:autoresearch automatically improves plugin skills by proposing mutations, evaluating them deterministically, and keeping or reverting based on results.
AI-generated summary based on this skill's SKILL.md
Decision gist · record as of 2026-07-27
lab:autoresearch automatically improves plugin skills by proposing mutations, evaluating them deterministically, and keeping or reverting based on results. lab:autoresearch runs a self-improving loop that mutates plugin skills one change at a time, scores each against deterministic criteria, and commits improvements while reverting failures. Target the weakest skill or run a full sweep overnight with /loop.
Use it when
- Yes.
- lab:autoresearch detects and reverts skill regressions while keeping successful improvements.
Verify before relying
Read SKILL.md below before installing (13 files). Open directory: indexed for reading, not audited.
Install
oliver-kriska/claude-elixir-phoenix/autoresearch · repository language: Python
Open directory. Skills are indexed for reading, not audited. Review a skill's body before installing it.
Frequently asked questions
AI-generated answers based on this skill's SKILL.md and metadata
How does lab:autoresearch auto improve plugin skills?
lab:autoresearch automatically improves plugin skills through iterative mutations and deterministic evaluation. It mutates skills one change at a time, scores each mutation against fixed criteria, commits improvements that pass, and reverts failures. You can target the weakest skill dimension or run a full sweep overnight using the /loop command.
Can I run overnight skill optimization loops with lab:autoresearch?
Yes. lab:autoresearch supports running overnight skill optimization loops with targeted or sweep strategies. Use /loop to launch an optimization run—target a specific weak dimension for focused improvement or execute a full sweep across all skill aspects. The loop runs deterministic evaluation on each mutation and commits only successful changes.
Does lab:autoresearch detect and revert skill regressions?
lab:autoresearch detects and reverts skill regressions while keeping successful improvements. Each mutation is scored against deterministic criteria; if a change degrades performance, it's automatically reverted. Only mutations that pass evaluation are committed, ensuring your skills improve without degradation.
How does lab:autoresearch score and track skill weaknesses?
lab:autoresearch scores and tracks the weakest skill dimensions for focused optimization. It evaluates each skill against deterministic criteria and identifies which aspects need the most work. You can then use targeted /loop runs to concentrate improvements on those specific weak areas rather than optimizing blindly.
What is automated skill mutation testing in lab:autoresearch?
lab:autoresearch performs automated skill mutation testing by making one change at a time to your plugin skills, evaluating each mutation deterministically, and keeping or reverting based on results. This iterative skill enhancement approach builds a mutation history and maintains a backlog of deferred optimizations for continuous refinement.
SKILL.md
Rendered from the published skill. Quoted content, verbatim.
Autoresearch — Plugin Skill Self-Improvement
Iteratively improve plugin skills via the autoresearch pattern: propose one mutation -> eval -> keep/revert -> repeat.
Usage
/lab:autoresearch # Targeted: attack weakest skill+dimension
/lab:autoresearch --skill review # Focus on one skill
/lab:autoresearch --strategy sweep # Process all skills alphabetically
/lab:autoresearch --dry-run # Show what would change, don't commit
For overnight runs:
/loop 5m /lab:autoresearch --strategy sweep --max-iterations 200
Iron Laws
- ONE mutation per iteration — if description needs "and", split into two
- NEVER mutate read-only files — check program.md before
(truncated - see the full file via the links below)
File tree — 13 files
lab/autoresearch/.gitignore
lab/autoresearch/SKILL.md
lab/autoresearch/program.md
lab/autoresearch/references/mutation-strategies.md
lab/autoresearch/references/state-management.md
lab/autoresearch/retention.py
lab/autoresearch/scripts/checks.sh
lab/autoresearch/scripts/protected_sections.py
lab/autoresearch/scripts/run-iteration.py
lab/autoresearch/scripts/score-skill.py
lab/autoresearch/tests/__init__.py
lab/autoresearch/tests/test_deviation_dispatch.py
lab/autoresearch/tests/test_protected_sections.py
Let your AI agent find skills like this
Example. Real query, live index.
You found this page by searching. An agent finds it by wishing: SkillFed indexes 56,283 agent skills by what they can do, searchable in plain language.
wish › “Automatically improve plugin skills through iterative mutations and deterministic evaluation”
Give your agent the search over MCP, or paste the wish link into any chat. No install? Search from any chat →
Related skills
Autoresearch automates large-scale SKILL.md improvement by spawning multiple variants per generation, evaluating each through a programmatic harness without browser overhead, and scoring with triple-run averaging to identify the strongest performer. The engine iterates autonomously, committing improvements to git and halting when scores plateau or oscillate.
Auto Paper Improvement Loop runs 2 rounds of autonomous review-fix-recompile cycles to enhance paper quality. Each round uses fresh LLM review to catch structural issues, theoretical inconsistencies, and presentation gaps, then implements fixes and recompiles. Optional style reference and edit whitelist let you steer structural choices or constrain which files the loop modifies.
self-learn captures patterns, failures, and decisions from substantial sessions, scores them for generality and confidence, then proposes targeted framework updates—from project-specific steering rules to cross-cutting framework changes. It's the mechanism that lets your framework evolve from real usage, surfacing diffs for approval before any changes apply.
Skill Developer is a meta-skill for building new Claude Code skills from the ground up or converting existing scripts. It provides templates, step-by-step guidance for wrapping MCP pipelines, and configuration patterns to enable Claude's automatic skill discovery.
Skill Generator is a meta-skill that scaffolds new Claude Code skills with your choice of execution model. Select sequential mode for linear pipelines with fixed dependencies, or autonomous mode for dynamic, context-aware routing. It generates a complete skill package including directory structure, phase files, and specification templates.
Extract patterns from recent sessions and convert them into persistent capabilities—rules for heuristics, skills for workflows, and hooks for automation. The skill guides you through gathering learnings, consolidating similar patterns, detecting meta-themes, and proposing artifacts with appropriate signal thresholds before creation.