agentleFS
Sign inStart free
ThakurvaibhavThakurvaibhav/platform-agent-harness/skills/ingest-reading/SKILL.md

ingest-reading

Skill2 starsChanged 7 days agoApache-2.0View on GitHub

Ingest an external source (web URL or local file) into the reading-notes corpus at ~/.agent-knowledge/reading/. Use when the user shares an article/paper/talk to "add to learnings/reading", "save this for later", "ingest this url", "distill this into a note", or drops a file to capture. Fetches, distills into a durable vetted note (NOT a raw dump), gets the user's approval interactively, then writes it and registers it in the reading index. Do NOT use for project learnings (those go in references/learnings-*.md via consolidation) or for answering questions.

The file

We split this skill into the rules it gives an agent

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

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    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
  1. Distinctive:

    If it's paywalled/JS-heavy and fetch is thin, tell the user and ask for a paste/PDF.

    line 26
  2. Distinctive:

    If it doesn't apply directly, say so — "useful as background" is a valid outcome.

    line 32
  3. Distinctive:

    Ask them to approve / edit / cut sections / reject.

    line 39
  4. Very specific:

    Cross-link back (optional, if strong): If the note materially informs a project learning, add a one-line pointer in that learnings-*.md …

    line 48
  5. Distinctive:

    Never write without the user's approval (step 3).

    line 57

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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.

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

Ingest Reading

Turn an external source into a durable, vetted reading note in the shared knowledge base. Model: Karpathy LLM-wiki — compile the source into synthesized knowledge, don't store raw text. The corpus is a separate trust tier from project learnings: external, curated, not yet proven in your stack. Keep dirty/unvetted info out by running the vet step interactively.

Corpus: ~/.agent-knowledge/reading/ · template: reading/_template.md · index: reading/index.md

Steps

  1. Get the source. URL → fetch (WebFetch, or gh api for GitHub). Local file → read it. If it's paywalled/JS-heavy and fetch is thin, tell the user and ask for a paste/PDF.
  2. Distill (do not dump). Produce a candidate note using reading/_template.md:
    • TL;DR — the one durable idea.
    • Key takeaways — durable points, each usable without re-reading the source.
    • How this applies to us — your platform lens (k8s/istio/argocd/agents/etc.). If it doesn't apply directly, say so — "useful as background" is a valid outcome.
    • Frontmatter: title, source_url, author, date_added (today), tags, maturity (external-opinion | established-practice | reference), applies_to, status: draft.
    • Cross-links: grep ~/.agent-knowledge/references/learnings-*.md (or run knowledge-search.sh <topic>) for related project learnings; list real [[learnings-x]] links.
  3. Vet interactively (the dirty-info guard). Show the candidate note to the user. Ask them to approve / edit / cut sections / reject. Apply their edits. Do NOT write until they approve. If they reject, stop — nothing is written.
  4. Write. Slug the title (kebab-case, dated if useful) → write to ~/.agent-knowledge/reading/<slug>.md. Set status: vetted.
  5. Register. Append a row to the ## Notes table in reading/index.md (Note link · Source · Tags · Applies to · Maturity). Remove the placeholder row if present.
  6. Cross-link back (optional, if strong). If the note materially informs a project learning, add a one-line pointer in that learnings-*.md (e.g. under a relevant entry: "See also reading/<slug>.md"). Keep it sparing — only genuine connections.
  7. Confirm. Tell the user the path written + the takeaways captured. If the reading corpus is large and search felt slow, remind them the qmd flip is available (see reading/index.md).

Rules

  • Never write without the user's approval (step 3). Never dump raw article text — distill.
  • One note per source. If re-ingesting an updated source, update the existing note in place.
  • Reading notes are FILES, not bd memories. The hive stays operational; reading is durable prose.
  • Notes live only in ~/.agent-knowledge/reading/ — do not touch references/learnings-*.md content except the optional one-line cross-link in step 6.

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.

Check your setup, free