Set up
Skills are installed by placing their folder where your agent loads skills.
The download comes from github.com/kwcantrell/okf-bundles itself, not from our server, so what you get is exactly what its author published.
Install instructions come from the README
README
synced Oct 23, 2026okf-bundles
A collection of Open Knowledge Format (OKF) bundles, plus a Claude Code router skill that makes them discoverable to agents.
OKF is a simple pattern for durable, sourced knowledge: Markdown files with YAML frontmatter, cross-linked into a graph, readable by both humans and AI agents without any special tooling.
Bundles
oks/git-best-practices/— Git and GitHub best practices, from the repository model to collaboration workflows.oks/ai-agent-repo-structure/— how to structure a repository so AI coding agents can discover context, load skills, and work reliably.oks/claude-best-practices/— how to use Claude Code effectively, from managing context and planning to verification, subagents, automation, and permissions/safety.oks/autoresearch-best-practices/— Karpathy's autoresearch autonomous ML-research loop, running it with Claude Code, and the agentic-engineering discipline around it.oks/agentic-sdlc-best-practices/— SDLC best practices when AI agents do the engineering: the vibe-coding→agentic-engineering spectrum, context engineering, phase-by-phase practices, harness engineering, and procedures for authoring AGENTS.md files, system prompts, and agent tooling.
Router skill
.claude/skills/oks-bundles/ is a
Claude Code skill that routes questions to the right bundle. Its
description carries explicit triggers (git/GitHub topics, or repo layout
for AI agents) so Claude Code loads it automatically when relevant, and its
body gives a decision table pointing at each bundle's index.md.
How to consume a bundle
Start at a bundle's index.md and follow its links — each index links to
area indexes, which link to individual concept files (progressive
disclosure). Every concept file ends with a # Sources section citing the
primary documentation behind its claims, so you can verify or go deeper.
To check structural integrity (frontmatter, required type field, link
resolution), run the validator:
python3 tools/validate_okf.py
With no arguments it checks every concept under oks/ plus the cross-references
in README.md and the router skill; pass a bundle path (e.g.
tools/validate_okf.py oks/git-best-practices) to check just that bundle. The
same command runs in CI on every push and pull request
(.github/workflows/validate.yml).
Design docs
See docs/superpowers/ for the design spec and
implementation plan behind this repo.