Analysis updated 2026-05-18
Teach Claude Code or Codex to author OKF knowledge bundles for datasets and APIs.
Validate whether an existing OKF bundle conforms to the v0.1 specification.
Give an AI agent a quick reference to the OKF format instead of writing custom prompts.
| lorsabyan/okf-skill | 0marildo/imago | 0xdfi/glm-5.2-1m-4x-dgx-spark | |
|---|---|---|---|
| Stars | 3 | 3 | 3 |
| Language | Python | Python | Python |
| Setup difficulty | easy | easy | hard |
| Complexity | 1/5 | 2/5 | 5/5 |
| Audience | developer | general | researcher |
Figures from each repo's GitHub metadata at analysis time.
Just copy the okf folder into the Claude Code or Codex skills directory.
okf-skill is an Agent Skill, a small package of instructions and tools that teaches coding assistants like Claude Code and OpenAI Codex how to work with a format called Open Knowledge Format, or OKF. OKF bundles are directories of markdown files with YAML frontmatter that describe things like datasets, tables, metrics, APIs, and playbooks. Instead of writing that documentation by hand, this skill gives an AI agent the knowledge it needs to author, validate, and read these bundles correctly. The repository is organized around a single skill folder. Inside it, a file called SKILL.md holds a quick reference to the OKF format along with workflows for authoring, validating, and consuming bundles. A references folder contains the full OKF v0.1 specification, copied from Google's knowledge-catalog project. A scripts folder holds a validator written in plain Python with no external dependencies, which checks whether a given bundle follows the format correctly. Installing the skill is a matter of copying or symlinking the okf folder into the right place. For Claude Code, that means placing it in a personal skills directory that applies across all projects, or in a project-specific skills folder. Once it is there, Claude Code recognizes tasks related to OKF or knowledge bundles automatically, or a user can call it directly with a slash command. Codex CLI uses the same file format, so the folder can be copied into its skills directory in the same way. The included validator script can be run from the command line against any bundle directory. It checks required structural rules, such as parseable frontmatter and non-empty type fields, as hard errors, and flags softer issues like broken cross-links or missing descriptions as warnings. A strict mode turns those warnings into errors as well. The author reports that the validator passes cleanly against the three reference bundles shipped with the original OKF specification project. The repository also points to a companion project called okf-reader, a separate web app for browsing OKF bundles. The skill itself is released under the Apache 2.0 license, though the reproduced specification document keeps its original Google copyright.
An Agent Skill that teaches Claude Code and Codex to author, validate, and read Open Knowledge Format documentation bundles.
Mainly Python. The stack also includes Python, Markdown, YAML.
You can use, modify, and distribute this project freely, including for commercial purposes, as long as you keep the copyright and license notices.
Setup difficulty is rated easy, with roughly 5min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.