agent-skills-platform
Build tested agent skills and govern their lifecycle through a user-defined marketplace: evidence, discovery, updates, rollback, quarantine, and 17-platform dis
What is agent-skills-platform?
Build tested agent skills and govern their lifecycle through a user-defined marketplace: evidence, discovery, updates, rollback, quarantine, and 17-platform distribution.
Overview of agent-skills-platform
What it is
Agent Skills Platform is an open-source tool that turns real workflows into tested, installable agent skills. It allows users to provide a prompt plus evidence (spreadsheets, reports, emails, screenshots, transcripts, links, or scripts) and produces a reusable workflow package that includes instructions, functional scripts, evaluations, security checks, a representative-run record, and a correction path. The platform also provides a governed marketplace for skill lifecycle management: evidence, discovery, updates, rollback, quarantine, and distribution across 17 platforms.
Key features
- Builds a reusable skill package from a prompt and supporting evidence, without requiring code from the user.
- Automatically runs Semantic Recon before implementation to create a pinned data contract when the workflow uses APIs, databases, MCP, codebases, or structured files.
- Validates every skill as one connected system using a skill graph that links instructions, scripts, evaluations, and expected outputs.
- Runs four parallel checks: specification, pipeline, security, and evaluation schema.
- Requires a representative run to prove the skill produces a useful result before publication.
- Supports governed team marketplaces with ownership, approvals, versioned releases, discovery, rollback, quarantine, and compatibility evidence.
- Can certify skills from real agent runs using
team_marketplace.py reliability(Caliper).
Use cases
- Converting a monthly revenue-variance review into a reusable internal skill that preserves expert judgment.
- Governing the lifecycle of agent skills within a team or organization, including approval, distribution, updates, and rollback.
- Preserving expert knowledge by capturing the question, evidence, decision, and success measure behind recurring work.
- Ensuring trust in shared skills through validation, security checks, evals, and representative runs before publication.
Getting started
See the repository README for installation instructions.
License and pricing
Licensed under MIT. The project is open source with no stated pricing.
Based on the project's public README and repository metadata.