- Home
- Skills
- Agents & Orchestration
- llm-wiki
Llm Wiki 1
Maintain a structured, compounding Markdown wiki that synthesizes knowledge across multiple documents and sources.
Free
Works with the AI Agents you already use
llm-wiki
Example session with this skill installed
Ingest the new research paper at raw/papers/attention.txt into the wiki and update relevant concepts.
- Read your context and instructions
- Compiled the llm-wiki
Processed the source and compiled it into the wiki — 1 new source page, 2 entity pages, 2 concept pages, with provenance recorded.
📄 Reading source: attention.txt
🔍 Pass 1/4 — Extracting entities, concepts, and claims...
Title : Attention Is All You Need
Entities : ['Google Brain', 'Ashish Vaswani']
Concepts : ['Transformer architecture', 'Self-attention']
📝 Pass 2/4 — Writing source page: wiki/sources/2026-11-10_attention-is-all-you-need.md ✓
👤 Pass 3/4 — Updating 2 entity page(s)...
✓ wiki/entities/google-brain.md
✓ wiki/entities/ashish-vaswani.md
💡 Pass 4/4 — Updating 2 concept page(s)...
✓ wiki/concepts/transformer-architecture.md (status: stub → developing)
✓ wiki/concepts/self-attention.md
📚 index.md updated · log.md updated · 🔗 provenance recorded (SHA-256)
✅ Ingest complete — Pages written: 5 · Entities: 2 · Concepts: 2
Next: ask a question ("How does self-attention work?") to get a cited answer compiled from these pages, or run lint after ~10 ingests to catch contradictions and orphans.
Connects securely to your tools. The creator never sees your data.
About this skill
Most AI systems rely on RAG, re-discovering information from scratch every time you ask a question. This leads to fragmented answers, missed connections, and inconsistent summaries across different chat sessions.
This skill implements a compiled knowledge base that treats information as a durable asset. Instead of volatile retrieval, it transforms raw documents into a structured, interlinked Markdown wiki. It builds a "second brain" where concepts, entities, and contradictions are synthesized once and maintained over time.
What it does
- Source Compilation generates strictly templated source pages with summaries, key claims, and provenance tracking.
- Entity Management maintains factual records for people, organizations, and systems, using aliases to prevent duplication.
- Concept Synthesis aggregates ideas across multiple documents, identifying tensions where sources disagree.
- Persistent Querying files valuable research answers as standalone pages with inline citations to the wiki.
- Integrity Linting performs periodic health checks to find orphaned pages, stale links, and index drift.
How it works
- INGEST new raw documents by creating source pages and updating all affected entity and concept files.
- QUERY the compiled wiki by reasoning over synthesized concepts rather than raw text chunks.
- MAINTAIN the master
index.mdandlog.mdto ensure every new piece of information is cross-referenced. - LINT the directory structure to resolve contradictions and update page maturity statuses.
Frameworks & tools
This skill is tool-agnostic and operates on standard Markdown files with YAML frontmatter. It is optimized for environments like Claude Code or Cursor that can navigate local directory structures and manage file hierarchies.
Why this beats prompting it yourself
Generic prompts often produce shallow summaries that exist only in the current context window. This skill enforces a rigorous architectural pattern—source, entity, concept, and query layers—ensuring that every new document actually improves the total knowledge of the system rather than just adding noise.
Use cases
- Research Synthesis for tracking academic papers where findings overlap or conflict.
- Technical Documentation for internal systems where entities and dependencies change frequently.
- Meeting Intelligence to turn transcripts into a persistent map of project decisions and stakeholders.
- Competitive Analysis for building a structured view of market players and industry frameworks.
Known limitations
Does not support automatic file deletion; outdated pages are marked as deprecated. Requires a specific local directory structure to function correctly.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Agents & Orchestration.
Requires a specific local directory structure. Does not support auto-deletion of pages (uses deprecation). Reasoning is limited to the compiled wiki contents.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 6 days ago
- Free with an Agensi account
Requires
Nothing to install.
Permissions
File Scopes
Reads source documents and reads/writes compiled markdown pages only. No shell, network, or external services — the agent executes the methodology with its own file read/write tools.