daikichiba9511/dotfiles/llm-wiki-ingest
- community
Checking that the SKILL.md file is still in the repository.
Use when the user asks to add a provided source or files under raw/ to an existing LLM Wiki and update source-backed pages and indexes. Do not use for ordinary summarization or querying existing wiki knowledge.
Install
npx vskill@latest install daikichiba9511/dotfiles/llm-wiki-ingestSKILL.md
LLM Wiki Ingest
Integrate a source into an existing LLM Wiki: create a source summary, update related pages, maintain cross-links, update index.md, and append log.md. Read references/pattern.md for the underlying model. Read references/integration-brief.md before presenting an ingest plan. Read references/page-templates.md before creating pages.
Usage
$llm-wiki-ingest [source_path_or_url] [wiki_dir]
- With a local path: read the file and ingest it.
- With a URL: fetch the content, save a raw copy under
raw/, then ingest it. - With no source: detect unprocessed files in
raw/by comparingraw/withlog.md.
Wiki Detection
- Prefer an explicit
wiki_dirfrom the user's request. - Otherwise search for
AGENTS.mdfiles that mentionLLM Wiki. - Fallback: search for legacy
_schema.mdfiles that mentionLLM Wiki. - If multiple candidates exist, ask which wiki to use.
- If none exists, suggest
$llm-wiki-init.
Workflow
- Read the wiki
AGENTS.md,index.md, and recentlog.mdentries. - Load the source. For markdown with local images, read the text first and inspect only images that matter to the source's claims.
- Check source fit against the wiki's
Source Admission PolicyinAGENTS.md. If the source is noisy, duplicate, low-quality, off-topic, or outside an unclear scope, ask before ingesting unless the user explicitly requested it. - Extract the source title, date, author or publisher when available, main claims, entities, concepts, concrete examples, numbers, and possible contradictions.
- Present a short integration brief before editing when the source is large, ambiguous, high-impact, uncertain-fit, or likely to update more than five pages. Use
references/integration-brief.mdfor the format. If the user requested unattended ingestion, proceed with stated assumptions instead of stopping. - Create one page in
pages/sources/usingreferences/page-templates.md. - For each important entity or concept, update an existing page or create a focused new page. Ask before creating many pages at once.
- Update
overview.mdorpages/syntheses/when the source materially changes the wiki's top-level understanding or an evolving thesis. - Add cross-links between the source page and every materially related entity, concept, analysis, or synthesis page.
- If the ingest reveals a repeated workflow preference, naming convention, page category, tag convention, or source admission rule, propose an
AGENTS.mdschema update. - Update
index.mdrows for created or materially updated pages. - Append one
ingestentry tolog.md. - Report pages created, pages updated, contradictions found, schema updates proposed, and follow-up sources or questions.
Rules
- Treat
raw/as immutable after a source is stored there. - Treat
raw/as curated, not a dumping ground. Do not ingest uncertain-fit sources without user confirmation or an explicit assumption. - Preserve existing sourced claims. If new evidence conflicts, add a contradiction note with both sources.
- Do not fabricate citation details. Mark missing title, date, or author as unknown.
- Prefer small focused pages over large omnibus pages.
- Reuse existing tags from nearby pages before inventing new tags.
- Do not save copyrighted source text beyond a reasonable raw copy requested by the user; generated wiki pages should summarize and cite.
- Keep the user in the loop on emphasis and scope for sources where multiple valid interpretations would lead to different wiki updates.
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingprivate_dot_agents/exact_skills/llm-wiki-ingest/SKILL.md
Popularity
- 7-day trend score
- 0
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/daikichiba9511/dotfiles/llm-wiki-ingest)