ingest
Skill0 starsChanged 8 days agoMITView on GitHub
Process sources from the second brain inbox (00-inbox/) into the wiki. Archive the original to 10-raw/, write a source summary page, update every concept/entity page it touches with bidirectional links, then update index.md and log.md. Use when the owner says "ingest", "process the inbox", "file this", "compile this source", or drops a file and asks to add it to the brain.
90 other projects use a version of this.
Adopted by
Most recently taken up by
- akovfoj-lang/plainframetheir own version, first seen Nov 16, 2026
- rafaym1/scholarship-desktheir own version, first seen Nov 16, 2026
- ryanportfolio/Corewise.Academytheir own version, first seen Nov 12, 2026
- fjwood69/moritheir own version, first seen Nov 10, 2026
- pche-broken-artist/forge-of-thoughttheir own version, first seen Nov 10, 2026
- playingpossumm/loose-endstheir own version, first seen Nov 10, 2026
Used with
Projects with it also use daily-arxiv, exp-eval and 10 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
LambenthanSkilldaily-arxiv
每日从 arXiv 拉取新论文,过滤相关性,auto-ingest 高优先级论文,检测 SOTA 更新
Also in 5 projects that use this
LambenthanSkillexp-eval
实验判决门:Review LLM 独立评判实验结果 → 4 种判决路径 → 自动更新 claims confidence、ideas status、graph edges
Also in 5 projects that use this
phanghonghaoSkillposter
HTML/CSS A4 海报生成。从沙龙/活动 Markdown 文件生成科技风 A4 海报 PDF/PNG。Use when user mentions: poster, 海报, 海报生成, make poster, generate poster, 活动海报, 沙龙海报.
Also in 5 projects that use this
skyllwtSkillexp-pilot-eval
Pilot result evaluation — read pilot results, apply success criteria, update idea page (pilot_result, failure_reason if failed), generate PILOT_VERDICT_REPORT. Called by /ideate Phase 5 after /exp-pilot-run.
Also in 4 projects that use this
skyllwtSkillexp-pilot-run
Pilot experiment execution — read Pilot Spec YAML, write pilot code, run experiment(Confirm with the user before operation and require the applicant to conduct manual inspection), return results. Called by /ideate Phase 5. Does NOT modify wiki pages or judge pass/fail.
Also in 4 projects that use this
- +6 more from Lambenthan
- +1 more from skyllwt
Other versions90 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 90
- garrytan/gbrain★ 33ksame name
- skyllwt/OmegaWiki★ 1.7ksame name
- DAIR-AutoResearch/AutoSci★ 1.7ksame name
- PKU-DAIR/AutoSci★ 1.7ksame name
- skyllwt/AutoSci★ 1.7ksame name
- suboss87/FDEOps★ 940same name
- coreyhaines31/makerskills★ 916same name
- jason-effi-lab/karpathy-llm-wiki-vault★ 718same name
- ZimoLiao/scholaraio★ 589same name
- egregore-labs/egregore★ 293same name
80 more not listed
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 repository11 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- brain-reviewSkill
- captureSkill
- gap-finderSkill
- lintSkill
- projectSkill
- querySkill
- repairSkill
- second-brainSkill
- setupSkill
The file
We split this skill into the rules it gives an agent
12 rules. 0 are common, 6 are distinctive, and 6 are very specific to this project.
0
Common
Most of the 74 edited versions we read keep these.
6
Distinctive
This author's own choices. Worth a look.
6
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 16
Check the owner's check-in preference in about.md ("always" / "only for big sources" / "auto").
- Distinctive: line 19
Read the whole source. For PDFs, read every page.
- Distinctive: line 31
Look for state changes: does this source supersede a current fact somewhere (role, version, price, plan)?
- Distinctive: line 56
Update the top summary if understanding changed.
- Very specific: line 57
Add facts under the right section with the source cited: … ([[source-page]]).
- Very specific: line 61
Add [[source-page]] to the frontmatter sources: list, bump updated:, and move seed → growing when it now has real content from 2+ sources.
- Distinctive: line 63
Add the link both ways: page ↔ source page, page ↔ related pages.
- Very specific: line 76
Search index: run scripts/qmd-sync.sh (no-op without qmd).
- Very specific: line 79
Run python3 scripts/lint.py and fix any errors you introduced (broken links, missing index lines, bad frontmatter).
- Very specific: line 88
Never edit anything already in 10-raw/ or 40-journal/.
- Distinctive: line 89
No claim without a source page citation.
- Distinctive: line 90
Keep the author's caveats and uncertainty.
“Common” is measured: kept by at least 6 in 10 of the 74 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
Ingest: 00-inbox/ → 10-raw/ + 20-wiki/
The most important operation. A wrong page written today is a wrong answer in every query for months,
so check in before writing. Rules and page format: AGENTS.md §3 and §6.
0. Prepare
- Session-start reading done? (
90-system/about.md,index.md, tail oflog.md.) If not, do it now. - List the inbox:
ls -1 00-inbox/ | grep -v .gitkeep. If several items are waiting and the owner didn't name one, list them with a one-line guess at each and ask which to start with (default: oldest first). - One source at a time. For a batch request, still finish each source's full cycle before the next.
- Check the owner's check-in preference in
about.md("always" / "only for big sources" / "auto").
1. Read
- Read the whole source. For PDFs, read every page; for long sources, read in chunks and take notes as you go.
- Images: read the text first, then view only the images that carry information.
- Note: claims, numbers, named people/orgs/tools, definitions, dates, and the author's caveats.
2. Map against the wiki
- From
index.md, list candidate existing pages. For each key concept/entity, search before deciding "new page":qmd search "<term> <alias>" -c wiki -n 10(if qmd is installed; see AGENTS.md §5a), elsegrep -ril "<term>" 20-wiki/, and checkaliases:in frontmatter. For fuzzy overlap ("is this idea already covered under another name?") use a structuredqmd querywithintent:+vec:. - Classify each touched page: update (adds info), contradicts (conflicts with an existing claim), create (no page exists and the topic matters beyond this one source), or skip.
- New pages: create as
status: seedwhen the source only mentions the topic in passing. - Look for state changes: does this source supersede a current fact somewhere (role, version, price, plan)?
3. Check in (unless the owner set "auto")
Present this and wait:
Source: <title> (<type>, <author>, <date>)
Takeaways: 1) … 2) … 3) …
Update: [[page-a]] (adds X) · [[page-b]] (new data on Y)
Conflict: [[page-c]] says P, source says Q. I'll add a ## Contradictions section
Create: [[new-concept]] (seed) · [[person-z]] (entity)
Skip: <what you're leaving out and why>Apply the owner's corrections to the plan. If they say the source is out of scope: skip to step 4, then write a
source page with status: deprecated and a one-line reason. Nothing is deleted.
4. Archive the original
- Target:
10-raw/YYYY-MM-DD-short-title.<ext>(date = today, the ingest date). - Attachments first: move them to
10-raw/assets/and fix their relative paths while the file is still in00-inbox/. Once a file is in10-raw/it is immutable. - Then
mv -n "00-inbox/<file>" "10-raw/<new-name>"(-n= never overwrite). If the name is taken, add-2.
5. Write the wiki
- Source page
20-wiki/sources/<short-title>.mdfrom90-system/templates/source.md. Setraw:to the archived path. Fill "What this changes in the wiki" as you go. - For each page in the plan (read it before editing):
- Update the top summary if understanding changed.
- Add facts under the right section with the source cited:
… ([[source-page]]). - Replace state facts in place with a new
(as of YYYY-MM-DD). Append events. Remove nothing the source doesn't actually supersede. - Conflicts → a
## Contradictionssection: both claims, both sources, no silent merge. - Add
[[source-page]]to the frontmattersources:list, bumpupdated:, and moveseed→growingwhen it now has real content from 2+ sources. - Add the link both ways: page ↔ source page, page ↔ related pages.
- New pages from
90-system/templates/concept.md/entity.md: full frontmatter, summary, at least one inbound link from an existing page. - Typical scale: a substantial source touches 5-15 pages. If it touched fewer than 3, say why in the report.
6. Bookkeep
index.md: add new pages under the right section (- [[page]]: one-line summary (status, N sources)); refresh summaries/source counts for pages that changed meaningfully.log.md(append at the bottom):## [YYYY-MM-DD] ingest | <Source Title> - created: [[a]], [[b]] · updated: [[c]], [[d]] · raw: 10-raw/YYYY-MM-DD-title.ext- Search index: run
scripts/qmd-sync.sh(no-op without qmd).
7. Self-check, then report
Run python3 scripts/lint.py and fix any errors you introduced (broken links, missing index lines, bad
frontmatter). Pre-existing issues: mention them, don't fix them now.
Report to the owner:
- Pages created / updated (as links), contradictions found, open questions raised.
- Suggested next: related sources to find, a question worth asking with
query. - Suggest a commit:
git add -A && git commit -m "ingest: <title>".
Hard constraints
- Never edit anything already in
10-raw/or40-journal/. Never delete pages. - No claim without a source page citation. Mark your own inferences "(inference)".
- Keep the author's caveats and uncertainty. Don't round them away.
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.

