skills.rest
nichobbs
nichobbsCOMMUNITY
Explore all skills by this author →
GitHub

self-improve

Run a research-to-plan-to-execute tournament loop to improve a target benchmark metric.

Updated Apr 29, 2026
Search Tags:#benchmarking#git worktrees#guardrails#code optimization#autonomous improvement#tournament selection
One-click install
npx skills add https://github.com/nichobbs/lyric-lang --skill self-improve-nichobbs
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: self-improve
Source: https://github.com/nichobbs/lyric-lang/tree/main/.claude/skills/self-improve
Command: npx skills add https://github.com/nichobbs/lyric-lang --skill self-improve-nichobbs

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve?

It reduces engineering iteration time by autonomously proposing, implementing, and evaluating code changes to improve a measurable repository metric.

Core Features & Use Cases

  • Autonomous improvement lifecycle: Runs end-to-end loop of research, hypothesis planning, architecture/critic review, execution, and tournament selection without pausing for approval between iterations.
  • Tournament selection with re-benchmarking: Compares candidate changes against a baseline, merges only winners after confirmation runs, and archives losing branches.
  • Safety guardrails for benchmark integrity: Enforces sealed-file constraints and validates plan/result JSON schemas to prevent self-modifying benchmark behavior.
  • Resumable execution: Supports crash recovery and iteration-level progress tracking under a resolved topic/scoped workspace.

Quick Start

Ask for a self-improvement run by providing the target repository path and the benchmark command you want repeatedly executed inside that repository.

Frequently Asked Questions about self-improve

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I autonomously optimize code to improve repository benchmark metrics?▼

You can autonomously optimize code by running a research-to-plan-to-execute tournament loop that proposes, implements, and evaluates changes, merging only those that measurably improve your target benchmark metric.

What is tournament selection for code optimization and how does it work?▼

Tournament selection for code optimization compares candidate changes against a baseline benchmark, merges only winning branches after confirmation runs, and archives losing branches to ensure strict metric improvement.

How do I automate performance regression fixes without manual code review between iterations?▼

You can automate performance regression fixes by deploying an autonomous improvement lifecycle that runs end-to-end research, hypothesis planning, critic review, execution, and tournament selection without pausing for approval.

Can I use git worktrees to isolate autonomous code improvement changes in my repository?▼

Yes, autonomous code improvement relies on git worktrees to provide worktree-based isolation for each candidate change, ensuring branches are evaluated independently before tournament selection and merging.

How are benchmark integrity and safety guardrails enforced during autonomous code improvement?▼

Benchmark integrity is protected by enforcing sealed-file constraints that prevent self-modifying benchmark behavior, alongside strict JSON schema validation for all planner and executor communications.

Does autonomous code improvement support crash recovery and resumable execution?▼

Yes, autonomous code improvement supports crash recovery and resumable execution by tracking iteration-level progress under a resolved topic and scoped workspace, allowing paused or crashed tournament loops to restart cleanly.

Related Skills

Explore More Software Engineering Skills→
JamesPrialJamesPrial
3

go-testing-benchmarks

Create Go benchmarks with b.N loops, sub-benchmarks, and timing control.

Community
Intermediate
Patasse97Patasse97

merge

Merge the winning agent's branch into the base and archive losing branches.

Community
Advanced
webdriveriowebdriverio
11

Global Validation

Enforce global validation across codebases with repository-wide checks.

Official
Basic
r3dlexr3dlex
2

self-improve

Automate software optimization through tournament selection, benchmarking, and automated merging.

Community
Advanced