Singularity-Claude: Add Self-Evolving Skills to Your Claude Code Workflow
What It Does — A Recursive Evolution Engine for Skills
Claude Code skills are static. You write them once, and they stay exactly as they are—even when they fail, produce subpar output, or encounter new edge cases. There's no feedback loop, no way to know which skills are working well, and no mechanism to improve them over time.
singularity-claude solves this by adding a recursive evolution loop to your skills. It's a self-evolving skill engine that creates, scores, repairs, and crystallizes skills through autonomous recursive improvement loops. Every skill execution gets scored, low scores trigger automatic repair, and high scores lead to crystallization—locked, battle-tested versions you can trust.
Setup — Install in Under 60 Seconds
No external dependencies. No SmythOS, no OpenTelemetry collector, no Docker. Pure Claude Code.
# Add the marketplace
claude plugin marketplace add shmayro/singularity-claude
# Install the plugin
claude plugin install singularity-claude
Or install from source:
git clone https://github.com/shmayro/singularity-claude.git
cd singularity-claude
claude plugin marketplace add .
claude plugin install singularity-claude
Once installed, singularity loads automatically in every new Claude Code session.
How It Works — The Four Commands That Change Everything
1. /singularity-create — Generate Skills from Patterns
When you notice a recurring pattern but don't have a skill for it yet:
/singularity-create
This command gathers requirements, generates a SKILL.md file, registers the skill, and initializes scoring. It turns your manual workflows into reusable, improvable skills.
2. /singularity-score — Rate Every Execution
After every skill execution:
/singularity-score
This dispatches an assessor agent that rates the skill across 5 dimensions (0-100) and auto-updates maturity levels: draft → tested → hardened.
3. /singularity-repair — Fix Broken Skills Automatically
When a skill scores below 50 average across 2+ runs:
/singularity-repair
The system analyzes telemetry, rewrites weak dimensions, bumps the version, and re-tests the skill. No more manually debugging broken skills.
4. /singularity-crystallize — Lock Battle-Tested Skills
When a skill averages ≥90 across 5+ runs and handles edge cases:
/singularity-crystallize
This creates a Git tag + locks the version, giving you production-grade skills that are immutable and rollback-safe.
When To Use It — Your Skill Quality Control System
Use singularity-claude when:
- You have more than 10 skills and can't manually maintain them all
- Skills fail silently and you only discover issues weeks later
- Edge cases accumulate and your skills become brittle
- You want to identify your most valuable skills through usage data
- Team collaboration where skill quality varies between developers
The Evolution Loop in Practice
Here's what happens automatically:
- Execute a skill →
/singularity-scoreruns automatically - Score < 50 →
/singularity-repairtriggers automatically - Score ≥ 90 →
/singularity-crystallizebecomes available - Every step is logged for full auditability
This means your skills get better with use, not worse. The more you use a skill, the more robust it becomes.
Start Today — Create Your First Evolving Skill
# 1. Start a new Claude Code session — singularity loads automatically
# 2. Create your first skill
/singularity-create
Describe a recurring task you do manually. The system will turn it into a skill that evolves with every use.
No more growing piles of skills where some work great, some are mediocre, and some are silently broken. With singularity-claude, every skill has a feedback loop, every execution contributes to improvement, and your best skills get locked in as production-grade assets.





