wiki-discover

Finds, credibility-ranks, and clips web sources for an Obsidian wiki.

Updated Jul 16, 2026
One-click install
npx skills add https://github.com/ehartye/wiki-master --skill wiki-discover-ehartye
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: wiki-discover
Source: https://github.com/ehartye/wiki-master/tree/main/skills/wiki-discover
Command: npx skills add https://github.com/ehartye/wiki-master --skill wiki-discover-ehartye

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Researching a topic for a knowledge base means manually searching, deduplicating against what you already have, judging source credibility, and clipping pages — a slow, error-prone process that often re-collects known material or loses track of rejected candidates. ## Core Features & Use Cases - Gap-aware discovery: Reads existing clippings and wiki coverage first so searches target unanswered questions, not duplicates. - Five-lens research: Runs Academic, Technical, Applied, News/Trends, and Contrarian perspectives (in parallel where the host supports it) to diversify candidate sources. - Credibility scoring and clipping: Scores candidates with an explicit rubric, records declines with a 180-day TTL, and clips survivors via clip.mjs (or clip-pdf/docx/xlsx/confluence variants) with quality and topic metadata. - Use Case: Ask it to research "vector database consistency models"; it dedups against your vault, returns a ranked source list, clips the survivors into raw/clippings, and hands off to wiki-ingest. ## Quick Start Use wiki-discover to find and clip credible sources on the topic "CRDT conflict resolution" for my wiki.

Frequently Asked Questions about wiki-discover

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I find new sources for an Obsidian wiki without duplicating existing ones?▼

Run wiki-discover with a topic; it first collects known source URLs from raw/clippings and summarizes existing coverage, then searches only for material that fills the identified evidence gap. Duplicates and previously declined URLs are dropped before clipping.

How does wiki-discover rank the credibility of web sources?▼

It scores each candidate with an explicit rubric: points for peer-reviewed or official sources, recency, credentialed authors, and cross-perspective corroboration, minus points for promotional content. Scores map to high, medium, low, or reject tiers, and only kept tiers are clipped.

Can wiki-discover clip PDFs, Word files, or Confluence pages?▼

Yes, but not through clip.mjs. PDFs route to clip-pdf.mjs, Word files to clip-docx.mjs, spreadsheets to clip-xlsx.mjs, and authenticated Confluence Cloud pages to clip-confluence.mjs. Each accepts the same --quality and --topic flags.

What happens to sources that fail to clip or are rejected?▼

Rejected candidates are recorded with a reason in a decline log that expires after 180 days, preventing re-litigation. Failed, thin, or wrong-node clips are queued to .wiki-master/triage.jsonl for later disposition via wiki-triage.

Does wiki-discover write directly to the wiki vault?▼

The research perspectives are read-only and never write the vault. Only the clip scripts write, and only into raw/clippings, after an operation is opened with op-begin.mjs and later committed with op-commit.mjs.