Skip to main content

Universal Skills

PyPI version Python License

Version: 1.5.0

Universal Skills is a portable catalog of agent capabilities and dependency-ordered skill workflows. It supports Claude-compatible SKILL.md discovery and the agent-utilities provider ecosystem.

Catalog model

  • universal_skills/<domain>/<name>/ contains one atomic skill with one capability.
  • universal_skills/<domain>-workflows/<name>/ contains a pure DAG over atomic skills or exact MCP tools; workflows do not embed business logic.
  • Platform and package operations belong under agent-packages/agents/<package>/<module>/skills/ and are exposed through the package's agent_connector_sdk.connector entry point.
  • assets/ may contain nested templates, but those are package resources rather than canonical catalog entries.

The catalog includes development, infrastructure, finance, research, product, creative, documentation, web, system, operations, content, and data capabilities. The content and data domains include evidence-aware drafting, publication preflight, dataset profiling, deterministic data-quality validation, and data-dictionary authoring.

Connector packages use mcp-builder for SDK scaffolding and shared pipeline contracts.

Catalog quality and roadmap

Reproduce the catalog audit locally:

python universal_skills/skills/skill-catalog-auditor/scripts/audit_catalog.py \
  universal_skills --format markdown

The audit reports trigger quality, atomicity signals, workflow dependency problems, unbound components, placeholder content, team-configuration drift, broken local links, and taxonomy consistency. It is a structural lint, so safety and domain correctness still require human review.

Install

Install the package, then deploy discovered skills to supported agent tools:

pip install universal-skills
install-skills --all-detected --symlink

For local development:

git clone https://github.com/Knuckles-Team/universal-skills.git
cd universal-skills
pip install -e '.[test]'
install-skills --all-detected --symlink

The repository also provides install.sh and install.ps1 for package installation and skill deployment. Review targets and configuration changes before using broad installation options.

Use from Python

from pydantic_ai_skills import SkillsToolset

from universal_skills.skill_utilities import (
    get_skill_graph_path,
    get_universal_skills_path,
)

skills = SkillsToolset(
    directories=[get_universal_skills_path(), get_skill_graph_path()]
)

Discovery respects each skill's enable/disable environment flag. Package-owned skills are discovered from installed provider entry points by the universal installer.

Author a skill or workflow

Use the bundled builders rather than copying a legacy entry:

# Atomic skill: creates only explicitly requested resource directories
python universal_skills/skills/skill-builder/scripts/init_skill.py --help

# Pure DAG workflow: resolves atomic skills or explicit MCP tools
python universal_skills/skills/skill-workflow-builder/scripts/build_workflow.py --help

Before opening a change:

python scripts/gen_bumpversion.py
bump2version --config-file .bumpversion.cfg --allow-dirty --no-commit --no-tag patch
python scripts/check_atomicity.py --strict
pytest -q
pre-commit run --all-files

Security boundaries

Audit and discovery are read-only. Skills that publish, send, delete, rotate, restart, deploy, trade, persist data, or otherwise change external state must make that side effect explicit and obtain the authorization appropriate to the target. Network scripts must verify TLS by default, bound requests, redact secrets, and treat retrieved content as untrusted input.

License

MIT. See LICENSE.

Metadata

Release files for universal-skills 1.5.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for universal-skills 1.5.0
File Size Uploaded Provenance Actions
universal_skills-1.5.0.tar.gz 3.2 MB Without provenance Details for universal_skills-1.5.0.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for universal-skills 1.5.0
File Interpreter ABI Platform Actions
universal_skills-1.5.0-py3-none-any.whl Python 3 none any Details for universal_skills-1.5.0-py3-none-any.whl

Total release size: 7.0 MB

Release files / universal_skills-1.5.0.tar.gz

Download URL universal_skills-1.5.0.tar.gz
Size 3.2 MB
Tags Source
SHA-256 checksum
How to use checksums
40b56cc38a852bb230bc46d39f2d615d331fcb49db1fc7057bb5ef3843b8eaf3
BLAKE2b-256 checksum
How to use checksums
8e40b0b5f0ba21cbb07eadee05e0c128384102748d64104be37a71227430e634
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.14.4

Release files / universal_skills-1.5.0-py3-none-any.whl

Download URL universal_skills-1.5.0-py3-none-any.whl
Size 3.9 MB
Tags Python 3
SHA-256 checksum
How to use checksums
d35c80706d2db9eae9d853c6a9a7048a54903c87ee9c860d210074865fd6b8db
BLAKE2b-256 checksum
How to use checksums
c71e370a15331727b6c1bf218c1e31499e9ad0a0a873f24d0754d70cab487f89
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.14.4

Security for universal-skills 1.5.0

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Release history Release notifications | RSS feed

This release

1.5.0 This release

2 release files No provenance

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page