agentleFS
LuciowayLucioway/way-stack/skills/vault-ingest/SKILL.md

vault-ingest

Skill0 starsChanged 2 months agoMITView on GitHub

Use when user says 'ingest', 'save this to vault', 'remember this source', or drops an article/PDF/note to summarize. Reads source → writes structured page to vault → updates index.md + log.md. Karpathy LLM Wiki pattern.

  • Installs packages

4 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

5 rules. 0 are common, 4 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 4 edited versions we read keep these.

  • 4

    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
  1. Distinctive:

    Do NOT ingest conversation noise (casual Q&A, status checks)

    line 67
  2. Distinctive:

    Do NOT duplicate — grep vault first for existing entry on same topic

    line 68
  3. Distinctive:

    Do NOT fabricate — only content actually in source

    line 69
  4. Distinctive:

    If nothing qualifies: say "nothing durable to save" and exit

    line 70
  5. Very specific:

    Read $VAULT_PATH from ~/.claude/CLAUDE.md or probe ~/Workspace, ~/Desktop/Workspace, ~/vault in that order.

    line 74

“Common” is measured: kept by at least 6 in 10 of the 4 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

vault-ingest — Karpathy LLM Wiki Ingest

When to use

User says:

  • "ingest this [article/PDF/note]"
  • "save this to vault"
  • "remember this source"
  • "take notes on this"

Or drops a URL / file / long paste and expects it saved for later retrieval.

Process

  1. Office input — if the source is docx/xlsx/pptx/pdf (or similar), convert it first: anydoc <file> -o <file>.md (npm install -g @firecrawl/anydoc). Ingest the resulting .md; keep the original file in the raw folder (00_INBOX/ or the project's raw dir). No anydoc? Say so and fall back to the Read tool (PDF) or ask for text.
  2. Read source — file, URL, or pasted text
  3. Discuss takeaways with user in chat — confirm what's worth keeping vs noise
  4. Classify target folder:
    • Raw / triage-later → 00_INBOX/
    • Reusable solution across projects → 02_KNOWLEDGE/patterns/
    • Architectural decision with alternatives → 02_KNOWLEDGE/decisions/ (ADR-NNN)
    • Milestone retrospective → 02_KNOWLEDGE/retros/
    • Single fix / gotcha / learning → 02_KNOWLEDGE/learnings/
    • Stable reference doc → 03_REFERENCE/
  5. Write summary page — structured markdown with frontmatter:

    ---
    type: pattern | decision | retro | learning | inbox | reference
    date: YYYY-MM-DD
    tags: [tag1, tag2]
    source: <url or path>
    ---
    
    # <Title>
    
    ## Summary
    <2-3 sentences of the core takeaway>
    
    ## Key points
    - <point>
    - <point>
    
    ## Related
    - [[other-note|title]]
  6. Update index.md — add entry under the right section, format:
    - [[02_KNOWLEDGE/<type>/<slug>|<title>]] — <one-line hook>
  7. Append to log.md:
    ## [YYYY-MM-DD] ingest | <type> | <title>
    <one-line hook>. See [[02_KNOWLEDGE/<type>/<slug>]]
  8. Update related pages — if new page references existing entities/concepts, add backlinks there too

Boundaries

  • Do NOT ingest conversation noise (casual Q&A, status checks)
  • Do NOT duplicate — grep vault first for existing entry on same topic
  • Do NOT fabricate — only content actually in source
  • If nothing qualifies: say "nothing durable to save" and exit

Locate vault

Read $VAULT_PATH from ~/.claude/CLAUDE.md or probe ~/Workspace, ~/Desktop/Workspace, ~/vault in that order. Confirm with user if ambiguous.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.