Okf-skill – progressive disclosure and deterministic conformance scaffolders (github.com)

A new AI coding agent skill called "Okf-skill" has been announced, designed to enhance the capabilities of AI systems like Claude Code, OpenCode, and Codex in reading and writing Open Knowledge Format (OKF) bundles, a specification developed by Google. This skill enables agents to construct and maintain a structured knowledge base using markdown files with YAML frontmatter, streamlining the process of accessing technical documentation. By leveraging concepts from Andrej Karpathy's "LLM Wiki," this initiative aims to eliminate inefficiencies where agents typically reread entire documents when searching for context or creating new content. The significance of Okf-skill lies in its capability to deliver deterministic conformance, making interactions with project documentation much more efficient. Rather than forcing AI agents to decipher documentation rules, the skill provides pre-programmed scripts to handle both simple tasks and more complex checks. This structured approach not only conserves resources and reduces token usage but also ensures consistency in the documentation process. Users can install Okf-skill easily through command line tools, granting agents the ability to scaffold knowledge bundles and validate their conformance, which is crucial in maintaining robust, up-to-date project documentation.