Back to browse
GitHub Repository

A self-learning skill layer for Claude Code — distills skills from your real sessions, updates them as you work, and prunes the ones that stop getting used. No daemon, no benchmark.

14 starsPython

Autoharness – a self-learning, maintaining skill layer for Claude Code

by Tigerless_ailab·Jun 30, 2026·3 points·0 comments

AI Analysis

●●●BangerBig BrainZero to One

Self-pruning skill layer for Claude Code that merges duplicates and updates from real usage.

Strengths
  • Learns from actual sessions without separate data collection loops
  • Merges similar skills instead of accumulating near-duplicates
  • Validates skills through invocation rate, not artificial benchmarks
Weaknesses
  • Only works with Claude Code, not other AI coding assistants
  • 14 GitHub stars indicates very early adoption and unproven at scale
Category
Target Audience

Developers using Claude Code who want persistent, improving skills

Similar To

Claude Code plugins · HAL

Similar Projects

AI/MLMid

Self-Evolving Skill – empirical results from a 5-round experiment

Five-gate validation framework prevents skill knowledge drift, but experiments limited to one domain.

Big Brain
tiansenxu
103mo ago
Developer Tools●●Solid

Agnix – lint your AI agent configs (Claude.md, skills, MCP, hooks)

Makes agent configs first-class with 229 domain-specific rules, autofix, and LSP support — so a tiny syntax mistake stops being a silent failure. The cross-editor plugins and GitHub Action are the standout moves: lint in your IDE and enforce checks in CI. I want a clearer map of which rules target which toolchains, but the breadth of integrations is impressive.

Niche GemSolve My ProblemSlick
anotherCodder
114mo ago
AI/MLMid

A governance pattern for self-evolving AI skills

Claude Code Skill pattern paper—interesting theory, but unclear if it ships as a usable tool today.

Big Brain
tiansenxu
103mo ago