Knowledge Skill

Project Cortex - Structural Knowledge Injection, Semantic Search & Persistent Memory.

Commands

Documentation Commands

search_documentation

Search markdown documentation and references for specific topics.

Example:

@omni("knowledge.search_documentation", {"query": "trinity architecture"})

search_standards

Search coding standards and engineering guidelines in docs/reference/.

Example:

@omni("knowledge.search_standards", {"topic": "python linting"})

Semantic Search Commands

knowledge_search (alias: code_search)

Semantic search for code patterns and documentation in knowledge base.

Example:

@omni("knowledge.knowledge_search", {"query": "error handling patterns", "limit": 5})

code_context

Get LLM-ready context blocks for a query.

Example:

@omni("knowledge.code_context", {"query": "how to handle errors", "limit": 3})

Memory Commands

update_knowledge_base

Save knowledge entry for future retrieval.

Example:

@omni("knowledge.update_knowledge_base", {
    "category": "patterns",
    "title": "Error Handling Pattern",
    "content": "Use Result types instead of exceptions...",
    "tags": ["error", "python"]
})

search_notes

Search existing notes and knowledge entries.

Example:

@omni("knowledge.search_notes", {"query": "error handling", "category": "patterns"})

summarize_session

Summarize current session trajectory into structured markdown.

Example:

@omni("knowledge.summarize_session", {
    "session_id": "session-123",
    "trajectory": [{"step": 1, "action": "search", "result": "found 5 files"}],
    "include_failures": true
})

Knowledge Ops Commands

ingest_knowledge

Ingest or update project knowledge base.

Example:

@omni("knowledge.ingest_knowledge", {"clean": false})

knowledge_status

Check knowledge base status.

Example:

@omni("knowledge.knowledge_status")

ZK Search Commands (Reasoning-based)

zk_search

High-precision search using ZK bidirectional links (PageIndex-style reasoning).

Example:

@omni("knowledge.zk_search", {"query": "agent skills progressive disclosure", "max_results": 5})

How it works:

1. Direct keyword search in titles/tags
2. LLM-style reasoning via bidirectional link traversal
3. Results ranked by relevance and link distance

zk_toc

Get Table of Contents for LLM context (all notes overview).

Example:

@omni("knowledge.zk_toc", {"limit": 50})

zk_hybrid_search

Hybrid search combining ZK reasoning + Vector search fallback.

Example:

@omni("knowledge.zk_hybrid_search", {"query": "architecture MCP", "use_hybrid": true})

zk_stats

Get knowledge base statistics.

Example:

@omni("knowledge.zk_stats")

zk_links

Find notes linked to/from a specific note.

Example:

@omni("knowledge.zk_links", {"note_id": "architecture", "direction": "both"})

zk_find_related

Find notes related to a given note using ZK's --related flag.

Example:

@omni("knowledge.zk_find_related", {"note_id": "agent-skills", "max_distance": 2})

Core Concepts

Best Practices

• Search first: Before adding new knowledge, search for duplicates
• Use categories: Organize entries by category (patterns/solutions/errors/techniques)
• Add tags: Use consistent tags for better retrieval
• Include examples: Code examples improve AI understanding

Advanced

• Semantic vs Text Search: Use knowledge_search for semantic understanding, search_documentation for exact matches
• Batch Ingest: Call ingest_knowledge with clean=false for incremental updates
• Session Continuity: Use session_id to link related sessions