القائمة
Ingest sources into the wiki.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
Browser automation via agent-browser. Use when the user wants to browse, interact with, or inspect web pages.
Review agent-facing knowledge (rules, AGENTS.md, CONVENTIONS.md, README.md, domain docs) for staleness and accuracy. Use when the user asks to audit, clean, or prune knowledge. Also use when they're confused by a rule/doc that seems wrong — treat that as the starting point for a broader audit.
Reflect on session, update rules, commit with jj.
Adversarial design interview.
Wiki health checks.
Query the personal wiki.
| name | obsidian-wiki-ingest |
| description | Ingest sources into the wiki. |
Process sources into the persistent wiki at ~/Documents/wiki/.
~/Documents/wiki/
├── raw/ # Immutable source documents (user-placed, never modified by LLM)
│ ├── inbox/ # Staging — drop files/URLs here
│ ├── articles/ # Web articles, blog posts
│ ├── notes/ # Markdown, text files
│ ├── papers/ # PDFs
│ ├── videos/ # Video files
├── wiki/ # LLM-maintained knowledge base
│ ├── concepts/ # Concept & topic pages
│ ├── entities/ # People, places, things
│ ├── sources/ # Source summaries
│ ├── synthesis/ # High-level overviews
│ └── analysis/ # Answers, comparisons, explorations
└── SCHEMA.md
| Tier | Examples | Stance |
|---|---|---|
| High | Peer-reviewed papers, official docs | Trust unless contradicted |
| Medium | Known-author blogs, conference talks | Verify surprising claims |
| Low | Reddit/HN/forum posts, anonymous sources | Demand evidence |
Flag unbacked strong claims: > ⚠️ **Unverified claim** from [source-type]: description — no evidence provided.
Separate observation from interpretation. Check for confounding factors. Never silently drop claims.
Trigger: /skill:obsidian-wiki-ingest (processes raw/inbox/) or /skill:obsidian-wiki-ingest <path-or-url>.
raw/ subdirectorywiki_search for existing coverage, check wiki/sources/<slug>.mdread for files, web_fetch for URLs, yt-dlp for YouTube transcriptsTMPDIR=$(mktemp -d)
yt-dlp --write-auto-sub --sub-format vtt --convert-subs srt --skip-download \
--output "$TMPDIR/{id}.%(ext)s" {url}
sed '/^[0-9]/d; /^$/d; /-->/d' "$TMPDIR/{id}.en.srt" \
| sed 's/<[^>]*>//g' | uniq > "$TMPDIR/{slug}.md"
# ... process, then:
rm -rf "$TMPDIR"
[[wiki links]] for all cross-referenceslowercase-with-dashes> ⚠️ Contradicts [[page]]: description