gap-finder
Skill0 starsChanged 9 days agoMITView on GitHub
Find what the second brain doesn't know yet and turn it into a ranked reading list. Collects open questions, Gaps sections from saved answers, wanted (linked but missing) pages, thin single-source pages, seeds, contradictions and journal ideas, ranks them against the owner's current focus, and proposes concrete sources or searches. Use when the owner asks "what should I read next?", "what are my knowledge gaps?", "what's missing in my wiki?", "where should I dig deeper?", or "find sources for X".
6 other projects use a version of this.
Adopted by
Most recently taken up by
- hassanzouhar/market_analyticstheir own version, first seen Sep 24, 2026
- jurgendn/agent-skillstheir own version, first seen Sep 17, 2026
- tansuasici/claude-research-kittheir own version, first seen Jun 2, 2026
- seandavi/scriptoriumtheir own version, first seen May 20, 2026
- npow/claude-skillstheir own version, first seen May 8, 2026
- 45ck/research-literature-review-skillstheir own version, first seen Mar 28, 2026
Other versions6 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 6
- jurgendn/agent-skills★ 15same name
- seandavi/scriptorium★ 5same name
- 45ck/research-literature-review-skills★ 4same name
- npow/claude-skills★ 4same name
- tansuasici/claude-research-kit★ 4same name
- hassanzouhar/market_analytics★ 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.
The file
We split this skill into the rules it gives an agent
7 rules. 0 are common, 6 are distinctive, and 1 is very specific to this project.
0
Common
Most of the 6 edited versions we read keep these.
6
Distinctive
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
- Very specific: line 19
If the owner named a topic ("gaps on pricing"), keep only signals related to it.
- Distinctive: line 30
Keep the top 5-7. Drop the rest into a one-line "also noticed" list.
- Distinctive: line 36
If web search tools are available and the owner wants it, find 2-3 candidate URLs per gap and label them by type and credibility.
- Distinctive: line 60
Ask which gaps to act on, then:
- Distinctive: line 67
Read-only on the wiki. The only writes are the gaps report and the log line.
- Distinctive: line 68
Never capture or ingest without the owner choosing the gap and the source.
- Distinctive: line 69
Don't invent sources. Candidate URLs must come from an actual search or the owner.
“Common” is measured: kept by at least 6 in 10 of the 6 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Gap-finder: what to learn next
Turns scattered "we don't know yet" signals into a short, ranked plan. Proposes only. Capturing sources
and ingesting them happen through capture / ingest with the owner's OK.
1. Collect (mechanical)
python3 scripts/gaps.py # or --jsonThis lists signals with the pages they came from: query gaps, open questions, wanted pages (with reference
counts), unresolved contradictions, thin pages (one source), seeds, and recent journal ideas, plus the owner's
current focus from 90-system/about.md. Don't re-scan by hand what the script covers.
If the owner named a topic ("gaps on pricing"), keep only signals related to it. With qmd, widen with
qmd query $'intent: gaps and open questions about <topic>\nvec: <topic> unknowns' -c wiki -c outputs.
2. Cluster and rank
- Cluster: merge signals that point at the same missing knowledge (a wanted page plus an open question plus a query gap about the same thing are one gap). Name each cluster as a question.
- Score each cluster (high/med/low) on:
- Focus fit: does it serve the current focus or an active project? (weighs most)
- Leverage: how many pages/answers would improve (reference counts, thin
stablepages, contradictions blocking conclusions) - Cost: is one good source likely to close it?
- Keep the top 5-7. Drop the rest into a one-line "also noticed" list.
3. Propose sources
For each top gap, suggest how to close it:
- The kind of source that would settle it (primary paper, official docs, a practitioner's write-up, data, an interview), and 1-2 concrete search queries.
- If web search tools are available and the owner wants it, find 2-3 candidate URLs per gap and label them by type and credibility. Don't capture anything yet.
- Gaps the owner can answer themselves ("what's my take on X?") get a note: "answer in your journal or tell me".
4. Save and log
Write 50-outputs/queries/YYYY-MM-DD-gaps.md (type: query):
# Knowledge gaps: YYYY-MM-DD
**Focus:** <from about.md>
| # | Gap (as a question) | Evidence | Why it matters | How to close it |
|---|---|---|---|---|
| 1 | … | [[page-a]] open question, [[query-x]] gap, [[wanted-page]] ×3 | unblocks … | primary source on …; search: "…" |
## Candidate sources (if searched)
- Gap 1: [title](url): type, credibility note
## Also noticed
- …Append to log.md: ## [YYYY-MM-DD] gaps | top: <gap 1>, <gap 2> + - saved: …. Run scripts/qmd-sync.sh.
5. Hand off
Ask which gaps to act on, then:
- URLs →
capture(theningest, one at a time). - Owner's own answers → they write in the journal; pick them up in
brain-review. - Wanted pages that existing sources already cover → offer a
repair-style fix (create the page from what's in the wiki, cited).
Hard constraints
- Read-only on the wiki. The only writes are the gaps report and the log line.
- Never capture or ingest without the owner choosing the gap and the source.
- Don't invent sources. Candidate URLs must come from an actual search or the owner.
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.

