Skip to content

emanon312/darwin-cull

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

2 Commits
 
 
 
 
 
 

Repository files navigation

Darwin Cull — Universal Skill Elimination Engine

"Survival of the fittest" — 适者生存

An ecosystem-agnostic skill culling engine. Works with Hermes, OpenClaw, Claude Code, and any AI agent that uses a skills directory.

8 Universal Rules

# Rule Detection Safety
1 Functional Duplicate Same-purpose skills, keep best ✓ Auto
2 Domain Overload Too many in one category ✓ Auto
3 Platform Mismatch Requires unavailable HW/OS/service ✓ Auto
4 Quality Gate SKILL.md incomplete/broken Review
5 Deprecated/Abandoned Marked deprecated, stale >180d ✓ Auto
6 Superseded by Built-in Core functionality now native Review
7 Stale Experiment WIP/alpha untouched >90d Review
8 Zero-Reference Orphan No skill references it Review

Features

  • Ecosystem Auto-Detection — Hermes, OpenClaw, Claude Code, or custom
  • Configurable Rules — via cull-rules.yaml
  • Protected Skills — never cull critical skills
  • Graveyard Archive — reversible, with .manifest.json traceability
  • Duplicate Groups — pre-define known duplicates
  • Domain Thresholds — cap skills per functional area
  • Platform Awareness — detect unavailable hardware/services

Quick Start

# Copy config template to your skills directory
cp references/cull-rules.yaml <your-skills-dir>/cull-rules.yaml

# Trigger culling
# Just tell your agent: "淘汰 skill" / "cull skills" / "清理冗余"

Configuration

See references/cull-rules.yaml for the full config template with comments.

Key settings:

ecosystem: auto           # auto-detect your agent
protected:                # NEVER cull these
  - darwin-cull
duplicate_groups: {}      # pre-define duplicates
domain_thresholds:         # cap skills per domain
  news: 5
  default: 6

Relationship with Darwin Skill

  • darwin-cull: Eliminate weak/redundant skills (breadth reduction)
  • darwin-skill: Optimize surviving skills (depth improvement)
  1. Cull → reduce noise, keep only viable skills
  2. Optimize → improve quality of remaining skills

Installation

git clone https://github.com/emanon312/darwin-cull.git <your-skills-dir>/darwin-cull

License

MIT

About

Darwin Cull — Universal Skill Elimination Engine. 通用Skill淘汰引擎。8条通用规则,支持Hermes/OpenClaw/Claude Code等任意Agent生态。可逆墓地归档。首次执行397→119(70%削减)。

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors