Okf query

Answer a question using ONLY an Open Knowledge Format (OKF) knowledge base (a wiki/ bundle of Markdown + YAML frontmatter), with citations and honest gaps. Use when the user asks something their wiki / knowledge base / notes should answer, or says "according to my notes/KB", "what does my wiki say about X", "search my knowledge base", "what do I have on X". Navigates the indexes, cites concept ids, and flags what's missing instead of guessing.From its SKILL.md

Install

npx -y skills add thattimc/skills --skill okf-query

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
runs commandsInstructs the agent to run 3 commands, including `grep -ril "<term>" wiki` and 2 more.

SKILL.md

1.9 KB, 379 tokens by cl100k_base, as published. Nobody here has run it

Query an OKF knowledge base

Answer from the wiki, grounded and cited. If the repo has a CLAUDE.md, follow its query guidance; this skill states the same discipline so it works standalone.

How to navigate (progressive disclosure)

Start at wiki/index.md → the relevant domain index.md → the concept page. Follow the description fields and relative links rather than loading the whole tree.
Useful shortcuts:
- grep -ril "<term>" wiki — find pages mentioning a term.
- grep -rl "^type: Source" wiki — list pages of a type.
- grep -rl "<tag>" wiki — find by tag. Frontmatter is plain text, so grep works.
- Follow # Related / cross-links and backlinks for neighboring concepts.

Rules

Answer only from KB content. Do not fill gaps from training data.
Cite the concept id / path for each claim, e.g. papers/routenlp-... or a relative link.
If the answer isn't in the KB, say so plainly and offer to add it with the okf-ingest skill — don't guess.
Distinguish what the KB asserts from your own inference; keep them clearly separated.
For "what's related to X", report the actual cross-links/backlinks, not associations you imagine.

In Obsidian (for the user)

Mention that they can also use Obsidian's Search (Cmd-Shift-F, including property filters like tag:#… or [year:2026]), the graph view, and the backlinks panel — all of which read the same relative-markdown links the KB uses.

What ships with it: 1 file

106 B alongside SKILL.md

agents/

openai.yaml106 B