FrancyJGLisboa/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.
What it solves
It solves the problem of turning messy, real-world business workflows into standardized, tested, and governed "skills" that AI agents can execute. Instead of relying on fragile prompts, it allows organizations to capture expert judgment and distribute verified workflows across a team without bloating the agent's runtime context with historical logs.
How it works
Users provide a prompt and evidence (such as spreadsheets, emails, or reports) of a recurring task. The platform then builds a reusable skill package containing instructions, functional scripts, evaluations, and security checks.
For workflows involving APIs or databases, it uses "Semantic Recon" to create pinned data contracts. The resulting skill is then verified through a representative run and can be published to a governed marketplace where a marketplace operator manages approvals, versioning, and distribution to other team members.
Who it’s for
- Subject Matter Experts (SMEs): Who want to codify their expert judgment into reusable AI skills.
- Marketplace Operators: Who manage the distribution, security, and policy for AI skills within an organization.
- AI Agent Users: Who need to install and use approved, high-quality workflows for their daily tasks.
Highlights
- Governed Marketplace: Provides a system for ownership, approvals, versioning, and rollbacks for team-wide skill distribution.
- Skill Packages: Bundles instructions, scripts, and evals into a single installable unit rather than just a prompt.
- Separation of Concerns: Keeps maintenance evidence and draft patterns separate from the executable instructions to prevent runtime bloat.
- Automated Verification: Uses a skill graph to ensure every expected result is tested and multi-step workflows have reliable entry points.
Related
- Project
- Project
- Project
- Project