agentleFS
Agents4Academia-AIAgents4Academia-AI/UReKA/.claude/skills/web/SKILL.md

web

Skill9 starsChanged 5 months agoMITView on GitHub

Ingest a vetted web resource (Wikipedia article, blog post, tutorial, or docs page) into a credibility-scored source file — curated registries first, then WebSearch/WebFetch

13 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

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

  • 0

    Common

    Most of the 12 edited versions we read keep these.

  • 1

    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. Very specific:

    If $ARGUMENTS is empty, ask the user for a URL, a wiki "<topic>", or a search "<query>", then continue.

    line 15
  2. Distinctive:

    If dropped, tell the user why (one line) and stop.

    line 56

“Common” is measured: kept by at least 6 in 10 of the 12 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

Ingest a factual, verifiable, widely-used web resource into sources/web_<slug>.md, recording a credibility score so every downstream citation is auditable. Prefers curated registries (Wikipedia, official docs, canonical blogs) and falls back to WebSearch/WebFetch, scoring each candidate with resources_cli.py (a deterministic domain-tier prior) plus your own judgment for borderline cases. Normally invoked by /curriculum as a pull step, but can be run directly.

If $ARGUMENTS is empty, ask the user for a URL, a wiki "<topic>", or a search "<query>", then continue.

Layout

Output goes to the personal sources/ by default; when invoked by /curriculum, write to that course's per-course library course/<slug>/library/ instead (the caller passes the destination directory). Either way the filename is web_<slug>.md. Template: .claude/Templates/web_source.md. The engine is .claude/src/resources_cli.py (stdlib; needs no API key). The credibility keep-threshold is 0.60 by default.

Mode A — wiki "<topic>" (curated registry, preferred)

  1. Search Wikipedia and review the scored hits:
    sh .claude/src/pyrun .claude/src/resources_cli.py --wikipedia "<topic>" -k 5
    Each JSONL line has title, url, snippet, score, tier. Pick the best-matching article (confirm with the user if ambiguous).
  2. Fetch the full article content:
    sh .claude/src/pyrun .claude/src/resources_cli.py --wiki-page "<exact title>"
    Returns title, url, summary, extract (full plain text), references, score.
  3. Format & write (Step "Write" below) using extract as ## Content. Keep the substantive sections; you may drop "See also"/"External links" tails (note the omission). Wikipedia is canonical tier — it clears the threshold.

Mode B — <url> (a specific page)

  1. Score the domain first:
    sh .claude/src/pyrun .claude/src/resources_cli.py --score "<url>"
  2. Fetch the page with WebFetch (returns clean Markdown). While reading, note the signals that refine credibility: a named author/byline, a references/citations section, and the publication year.
  3. Decide (see Credibility decision below). If kept, format & write. If dropped, tell the user why (one line) and stop.

Mode C — search "<query>" (discovery)

  1. Search with WebSearch for the query (optionally seed from curated hubs: sh .claude/src/pyrun .claude/src/resources_cli.py --hubs "<area>").
  2. Score each candidate's domain with --score <url> <url> ....
  3. Present the top 3–5 as a list labelled tier · score · why (highest first), and ask which to ingest. For each chosen URL, continue as Mode B from step 2.

Credibility decision

For each candidate combine the deterministic score with the signals you observed:

  • tier == "blocked" → never ingest.
  • score ≥ 0.60 → keep.
  • score in [0.45, 0.60) → escalate: judge from the fetched content whether it's authoritative, factually grounded, and widely-recommended (named expert author, cites primary sources, official/maintained). Keep only if yes; record your one-line reason.
  • score < 0.45 → drop unless the user explicitly insists.

Record the final score, tier, and a one-line rationale in frontmatter. Never ingest content you can't attribute to a real, identifiable source.

Write

  1. Get today's date: date +%Y-%m-%d.
  2. Slugify the title to snake_case → <dest>/web_<slug>.md (<dest> = sources/ by default, or course/<slug>/library/ when /curriculum invoked this skill). If it exists, ask before overwriting.
  3. Write following .claude/Templates/web_source.md:
    • type: web_source, title, url, retrieved: <today>.
    • credibility: block with score, tier, rationale.
    • concepts_mentioned: the key concepts the page covers (for retrieval/links).
    • source_links: the canonical URL.
    • ## Content: the cleaned objective text (headings, lists, code, LaTeX preserved; nav/ads/boilerplate stripped). Faithful — no personal input, no invented facts; note any large omissions.

Wrap up

Report: the file written, its credibility score/tier, concepts extracted, and any candidates dropped (one-line reason each). If invoked directly, offer to run /collate <topic> to fold this source into a paper/concept page.

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.