graph-ingest
/graph-ingest
Use when the user brings an external source worth keeping — a URL, a paper, a tweet/thread, a blog post, a docs page — and wants it pulled into the second brain. Fetches the source into the git-ignored capture inbox, admits it to the committed wiki through the admission policy, then refreshes the knowledge graph so it joins the rest of your thinking. Trigger on "/graph-ingest", "ingest this url", "add this paper to my knowledge", "add this tweet/page to my knowledge", "capture this source".
- Author
- alirezarezvani ↗
- Repository
- gaios ↗
- Licence
- MIT
- Last repository change
- 06.06.2026
Assessment and security
- Files in this package
- 1
- Scripts
- 0
- HTML comments
- 0
- URLs in SKILL.md
- 0
No signals detected. Repository changed on 06.06.2026, without references to retired models or APIs.
MIT, open: use and modify while retaining author attribution.
What it does and when to use it
Use when the user brings an external source worth keeping — a URL, a paper, a tweet/thread, a blog post, a docs page — and wants it pulled into the second brain. Fetches the source into the git-ignored capture inbox, admits it to the committed wiki through the admission policy, then refreshes the knowledge graph so it joins the rest of your thinking. Trigger on "/graph-ingest", "ingest this url", "add this paper to my knowledge", "add this tweet/page to my knowledge", "capture this source".
Triggers from the description
Original author description
Use when the user brings an external source worth keeping — a URL, a paper, a tweet/thread, a blog post, a docs page — and wants it pulled into the second brain. Fetches the source into the git-ignored capture inbox, admits it to the committed wiki through the admission policy, then refreshes the knowledge graph so it joins the rest of your thinking. Trigger on "/graph-ingest", "ingest this url", "add this paper to my knowledge", "add this tweet/page to my knowledge", "capture this source".
How it works
- When to run
- The output (always this shape)
- Process
- Autonomy
- Guardrails (from CLAUDE.md) — CRITICAL
SKILL.md preview
--- name: graph-ingest description: Use when the user brings an external source worth keeping — a URL, a paper, a tweet/thread, a blog post, a docs page — and wants it pulled into the second brain. Fetches the source into the git-ignored capture inbox, admits it to the committed wiki through the admission policy, then refreshes the knowledge graph so it joins the rest of your thinking. Trigger on "/graph-ingest", "ingest this url", "add this paper to my knowledge", "add this tweet/page to my knowledge", "capture this source". --- # Graph Ingest Capture an external source into the second brain **through the graph.** This is the bridge between graphify's `add` and the gAIOS wiki loop: fetch → admit → graph. One command, three guarded hops, ending with the new knowledge cross-linked in `wiki/` and visible in the graph. ## When to run - A URL / paper / tweet / thread / page is worth keeping, not just reading once. - The user says "save this", "add this to my knowledge", or pastes a link and wants it remembered. - You spot a source referenced 2-3 times across sessions → suggest ingesting it so it stops being re-fetched. ## The output (always this shape) ``` ## Ingested: <source title> **Source** — <url> (fetched <date>) **Captured** — raw/<file> (git-ignored inbox) **Admitted** — wiki/<entry>.md ⏸ draft — awaiting your OK to commit - de-identified: <what was stripped/referenced-out, or "nothing flagged"> - lint: <wiki_lint.py result — pass / N warnings> - cross-links: <related wiki entries linked> **Graphed** — graphify-out/ refreshed; new node(s): <node names> - notable edges (EXTRACTED / INFERRED): <1-2 connections worth seeing> **Next** — review the draft entry → approve commit → open graph.html ``` ## Process 1. **Fetch into the capture inbox.** Run `graphify add <url> [--author --contributor]` — it pulls the source into `raw/` (the git-ignored capture inbox), then re-extracts. Use `--author`/`--contributor` only when you know who wrote it. **Verify:** the new file landed in `raw/` and nowhere else. If the fetch failed, report the error; do not hand a half-fetched file forward. 2. **Admit it to the wiki.** Hand off to `/wiki` (`references/sops/wiki-translate.md`) to ADMIT the capture: read any existing entry first then merge, apply the **admission policy** (de-identify — strip/reference-out any secret, PHI/PII, named-deal specific, or live financial figure), cross-link related entries, update `wiki/_index.md`, then run `python tools/wiki_lint.py`. **Gate:** lint must pass (no broken-link/IBAN/secret errors) before the entry is eligible to commit. Leave the entry as a **draft**; do not commit without the human's OK. 3. **Refresh the graph.** Run `/graph` (default scope = code + committed `wiki/` only) so the admitted entry joins the graph. **Verify:** `graphify-out/` regenerated and the new entry appears as a node; name 1-2 notable edges from `GRAPH_REPORT.md`, labelled EXTRACTED vs INFERRED so claims stay honest. 4. **Move the capture.** Per the wiki SOP, the processed file moves `raw/` → `raw/_archive/` (never deleted). 5. **Hand back.** Surface the output block, point at the draft entry and `graph.html`, and offer to commit once approved. ## Autonomy **L2 — drafts the wiki entry; admission and commit are human-gated.** Fetch and graph-refresh run automatically (reversible, local); the wiki entry stays a draft until the human approves the commit. Never commit de-identified content on the user's behalf without that OK. ## Guardrails (from CLAUDE.md) — CRITICAL - **Fetched external content lands ONLY in git-ignored `raw/`.** It reaches committed `wiki/` ONLY after de-identification + a passing `tools/wiki_lint.py`. (Guardrails #1, #2, #7) - **Never graph `raw/` directly.** `raw/` can hold raw PHI/PII/financials before de-identification; the graph's default scope is code + committed `wiki/` only — `raw/`, `.env`, `.tmp/` are hard-excluded. - **No secrets / PHI / confidential figures into `wiki/`** — reference the source, never transcribe sensitive specifics. The wiki is committed; treat it as public. - **Cite, don't invent.** Keep the source URL on the entry; rely on graphify's EXTRACTED / INFERRED / AMBIGUOUS labels rather than asserting unverified links. - `graphify-out/` is a git-ignored derived artifact — don't commit it.
Versions
The database has no other skills named graph-ingest.
Database snapshot: 13.11.2026 · package version 968131ac0261 (SHA-256 digest of the entire package)
Downloads in the last 30 days: 0 · ranking
Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.
Similar skills
- 100wiki-ingest
Use when adding a new source to a wiki — a paper, article, URL, file, transcript, or any document. One ingest may touch 10-15 wiki pages. - 100ingest
File a source (URL, file path, or pasted text) into raw/ with provenance preserved. Trigger on "ingest this", "process this source", "file this article", "save this transcript", "add this thread to the OS" (or run /founder-os:ingest). Also fires when the user shares a source… - 100second-brain
Operate as the user's personal "second brain" — a compile-not-retrieve knowledge base following Andrej Karpathy's LLM Wiki pattern. Use when the user wants to ingest/save an article, note, paper, or source into their knowledge base; answer a question from their accumulated wiki;… - 90ingest
Compiles one source into a knowledge base's wiki: decides first whether the file belongs in the wiki at all, then reads it fully, confirms the takeaways with the owner, and writes or updates the compiled page, the index, and the log, leaving the source untouched. Works in any… - 100wiki-ingest
Compile articles, documents, or notes into a structured wiki knowledge base. Use when user says 'ingest to wiki', 'compile to knowledge base', 'update wiki', 'wiki ingest', 'add this to wiki', or invokes /wiki-ingest. Supports single or batch ingest. Triggers: wiki, ingest,… - 100ingest
Ingest one source (PDF, web page, or local markdown/text file) into the vault's LLM wiki — capture it into raw/, read it at the right cost, discuss it with the user, then write source/concept pages, update index, log and hot cache, and verify. Use when the user asks to ingest,…