Skip to main content

Govern AI skill growth through intelligent merging of similar capabilities — no blocking, only consolidation. Built for LangChain, CrewAI, and custom agent systems.

Project description

AI Skill Governance

Govern AI skill growth through intelligent merging — not blocking.

Prevent skill bloat in AI agents by automatically detecting and merging similar capabilities while preserving functionality.

PyPI Python Versions License: MIT

Support

If this project is useful to you, consider supporting development:

Ko-fi Buy Me A Coffee

Why AI Skill Governance?

Modern AI agents can accumulate dozens or hundreds of skills. This leads to:

  • Context dilution
  • Slower reasoning
  • Conflicting behaviors
  • Maintenance nightmare

AI Skill Governance solves this by intelligently merging similar skills instead of arbitrarily limiting them.

Before / After

A support agent accumulates 12 skills over a few iterations. Many overlap.

Cluster Before After
Greetings greet, welcome, say_hi, hello_user greet
Refunds refund, return_payment, issue_refund refund
Lookups find_order, search_orders, get_order_by_id find_order
One-offs escalate, close_ticket escalate, close_ticket
Total 12 skills · ~840 tool-list tokens* 5 skills · ~350 tokens

58% fewer skills. ~58% smaller tool-list prompt. Same coverage, faster tool selection, fewer conflicting behaviors.

*Illustrative. Tool-list tokens depend on serialization format (OpenAI function-calling schema, Anthropic tool blocks, etc.). The skill-count reduction is deterministic; token savings scale roughly with it.

Installation

pip install ai-skill-governance

For faster fuzzy matching:

pip install "ai-skill-governance[fuzzy]"

Quick Start

from ai_skill_governance import SkillRegistry

registry = SkillRegistry(default_threshold=0.75)

@registry.register(description="Sends a warm greeting")
def greet(name: str):
    return f"Hello, {name}!"

@registry.register(description="Welcomes the user warmly")
def welcome(name: str):
    return f"Welcome, {name}!"

# Check for suggested merges
print(registry.suggest_merges())

# Perform merge (v0.1.1: includes signature compatibility check)
merged = registry.perform_merge("greet", "welcome")
print(merged)

# Persistence (new in v0.1.1)
registry.save("my_skills.json")
registry.load("my_skills.json")

Advanced Merging

The merger creates a smart combined function that:

  • Preserves the primary behavior
  • Augments with secondary skill when signatures match
  • Intelligently combines string/dict results
  • Maintains full traceability

CLI

# Show status
skill-gov status --threshold 0.8

# Analyze and suggest merges
skill-gov merge --threshold 0.75

# Auto-merge similar skills
skill-gov merge --auto

Configuration

  • default_threshold: Similarity cutoff (0.0–1.0). Default: 0.75
  • merge_mode: "suggest" (default) or "auto"

v0.3.0 New Features

Skill Graph Visualization

from ai_skill_governance import SkillRegistry, SkillGraph

reg = SkillRegistry()
# ... register skills ...

graph = SkillGraph(reg)
print(graph.to_ascii())           # Simple text view
dot = graph.to_dot("skills.dot")  # Graphviz format
data = graph.to_json("graph.json") # For web dashboards

Semantic Merging (Major Upgrade)

pip install "ai-skill-governance[semantic]"
from ai_skill_governance import SkillRegistry

reg = SkillRegistry()
suggestions = reg.suggest_merges(method="semantic")  # Uses embeddings when available

LangChain Integration

from ai_skill_governance.integrations.langchain import LangChainSkillGovernor

governor = LangChainSkillGovernor()
# Use @governor.register_tool on your LangChain @tool functions

CrewAI Integration

from ai_skill_governance.integrations.crewai import CrewAISkillGovernor

PyPI Release Notes

This package is fully prepared for PyPI publication with:

  • Modern pyproject.toml
  • Proper metadata and classifiers
  • Optional dependencies (fuzzy for rapidfuzz)
  • Comprehensive documentation
  • MIT License

Contributing

We welcome contributions that improve merging intelligence, add new similarity methods, or enhance the governance dashboard.

License

MIT License — see LICENSE file.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

ai_skill_governance-0.3.0.tar.gz (16.1 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

ai_skill_governance-0.3.0-py3-none-any.whl (16.4 kB view details)

Uploaded Python 3

File details

Details for the file ai_skill_governance-0.3.0.tar.gz.

File metadata

  • Download URL: ai_skill_governance-0.3.0.tar.gz
  • Upload date:
  • Size: 16.1 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.12.3

File hashes

Hashes for ai_skill_governance-0.3.0.tar.gz
Algorithm Hash digest
SHA256 62e20aab692eaf4c8457df0a184afa38fabefe7ab7722ca13c05981aecdcf0b7
MD5 0841cd880a3e08f8679c69ab77bcb8aa
BLAKE2b-256 0f4cf75f6056cc508d5a8c916891a237156a508d1e7ee64d2ce7cfb01ffb78f1

See more details on using hashes here.

File details

Details for the file ai_skill_governance-0.3.0-py3-none-any.whl.

File metadata

File hashes

Hashes for ai_skill_governance-0.3.0-py3-none-any.whl
Algorithm Hash digest
SHA256 8cda1803984ec949affee4c2520dce170a66fa8c37db9e307d2f8eb87d857327
MD5 10db3958a7c235006f3767edcfe2b7d6
BLAKE2b-256 32e2fbcbac373b0b93a534e26fe80cc4fe009d1be56476cf499211e41b14d81c

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page