Design and initialize a new AI agent skill from scratch with proper boundaries.
Audit an existing skill's trigger conditions and action permissions for conflicts.
Migrate reusable capabilities from old projects while keeping examples separate.
Validate a skill's scripts, references, and metadata before deployment.
| cheshiremew/meta-skills | 920linjerry-stack/capital-studio | adya84/ha-world-cup-2026 | |
|---|---|---|---|
| Stars | 16 | 16 | 16 |
| Language | Python | Python | Python |
| Setup difficulty | moderate | easy | easy |
| Complexity | 3/5 | 3/5 | 2/5 |
| Audience | developer | researcher | general |
Figures from each repo's GitHub metadata at analysis time.
Requires cloning into a specific user-level skills directory and creating symlinks or junctions if developing from a separate location.
Meta-skills is a toolkit for working with Codex agent skills, which are reusable instructions that help AI agents perform specific tasks. Rather than simply checking folder structures or metadata, it focuses on understanding what a user actually needs a skill to do, what its responsibilities should be, and where its boundaries lie before generating files or validating behavior. The repository covers several core activities: designing and initializing a new skill from scratch, generating and maintaining the required documentation files, auditing a skill's trigger conditions and action permissions, migrating reusable capabilities from existing projects while keeping example materials separate, and validating scripts, references, metadata, and the final observable results a user would see. It also checks whether a skill's scope is too narrow, too broad, or has conflicting entry points. Installation involves cloning the repository into a specific user-level skills directory that Codex automatically discovers. For developers who want to keep working in a separate location, the README describes creating a directory junction on Windows or a symbolic link on POSIX systems so the source repository remains the single source of truth. If Codex does not immediately show the skill, restarting it usually resolves the discovery issue. Usage is straightforward. You can explicitly call meta-skills with a description of what you want to create, refactor, audit, migrate, or validate. Codex can also invoke it implicitly when a request matches the skill's description. The repository includes Python scripts for initializing a new skill, generating UI metadata, and running structural validation checks, though the README notes that structural checks do not replace real behavioral verification. The project keeps core principles in a protected, read-only section by default. Only when a user explicitly requests changes to those principles can the core content, lock file, and validation invariants be modified. The repository currently has no open source license, meaning public visibility does not automatically grant permission to copy, modify, or distribute the code.
A Python toolkit that helps you design, create, audit, and validate reusable instructions for AI agents, ensuring they perform specific tasks effectively before generating files.
Mainly Python. The stack also includes Python.
No license is provided, so you do not have permission to copy, modify, or distribute the code even though it is publicly visible.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Don't trust strangers blindly. Verify against the repo.