retrieve
Skill9 starsChanged 5 months agoMITView on GitHub
Find the knowledge-base notes relevant to a topic — BM25 seeds expanded by following relevant wikilinks — returned as a list of files with a one-line reason each
3 other projects use a version of this.
Adopted by
Most recently taken up by
- synaptiai/agent-capability-standardtheir own version, first seen Nov 1, 2026
- khanhmoiii/claude-tutor-skillstheir own version, first seen Oct 25, 2026
- opendatadiscovery/odd-teamtheir own version, first seen Sep 3, 2026
Other versions3 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 3
- synaptiai/agent-capability-standard★ 7same name
- khanhmoiii/claude-tutor-skills★ 1same name
- opendatadiscovery/odd-team★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository14 other files it gives its agents.
- .claude/AGENTS.mdAGENTS.md
- .claude/CLAUDE.mdCLAUDE.md
- alphaxivSkill
- askSkill
- autoexploreSkill
- collateSkill
- curriculumSkill
- editSkill
- goalSkill
- notionSkill
- obsidianSkill
- tutorSkill
- webSkill
- zoteroSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 3 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 16
If $ARGUMENTS is empty, ask the user for a query, then continue.
- Distinctive: line 20
Get seed notes: Run the Python retriever (it builds/refreshes the index automatically.
- Very specific: line 36
Look at the note's [[wikilinks]].
- Distinctive: line 47
Keep the expansion focused: follow only genuinely relevant links, don't re-visit notes, and stop once nothing new and relevant is being …
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Find the notes in the local knowledge base relevant to $ARGUMENTS. The base
spans four directories — sources/ (objective paper text), notes/ (personal
notes + annotations), papers/ (paper pages), and concepts/ (concept pages) —
and retrieval covers all four. It works in two parts: a Python BM25 search gives
seed notes (no LLM), then you expand from those seeds by reading each note
and following only the wikilinks that look relevant to the query. Return the
resulting set of relevant notes — a list of files, each with a one-sentence reason
it's relevant. Do not synthesise or merge their contents.
If $ARGUMENTS is empty, ask the user for a query, then continue.
Steps
Get seed notes. Run the Python retriever (it builds/refreshes the index automatically; needs no API key):
sh .claude/src/pyrun .claude/src/retrieve_cli.py --retrieve "$ARGUMENTS"Each output line is
<file path>\t<title>. These seeds start your working set. (Optional:-k Nfor more/fewer seeds.)- Expand by following relevant wikilinks. Maintain a relevant set (notes
you'll return) and a frontier of notes to examine, initialised to the seeds.
While the frontier is non-empty:
- Take a note off the frontier and
Readit (skip if already visited). - Decide if it's relevant to $ARGUMENTS; if so, add it to the relevant set with a one-sentence reason why.
Look at the note's
[[wikilinks]]. For each, judge whether that linked note is likely to hold material relevant to $ARGUMENTS. Only for the ones you judge relevant, resolve them to files and add any not-yet-visited notes to the frontier:sh .claude/src/pyrun .claude/src/retrieve_cli.py --resolve "tool use" "planning"Output is
<link>\t<path|NONE>; skipNONE(no page yet — a knowledge-base gap) and skip notes already visited.- Keep the expansion focused: follow only genuinely relevant links, don't re-visit notes, and stop once nothing new and relevant is being found (in practice ~1–2 hops out from the seeds is plenty).
- Take a note off the frontier and
Return the relevant set. Present each note as a clickable file path with its one-line reason, e.g.
- papers/instructgpt.md — defines the RLHF pipeline the query asks about - concepts/rlhf.md — followed from instructgpt's [[RLHF]] link; explains reward modelling + PPO - sources/zotero_instructgpt.md — primary annotation backing the pipeline detailsList the files only — do not synthesise, summarise, or merge their contents.
Notes
- This skill needs no Anthropic API key: the Python side is pure BM25/lookup, and the wikilink-following judgement is done by you (Claude) reading the notes.
- It does not write or edit any files.
- Used directly when the user asks to find relevant notes, and as the gathering
step for the
/concept(synthesis) and/ask(question-answering) skills.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

