---

name: taxonomy description: > Extract Federated Taxonomy tags from text. LLM extracts candidates, deterministic validation filters to known vocabulary. Returns bridge tags for multi-hop graph traversal. allowed-tools: ["Bash", "Read"] triggers:

• taxonomy
• tag this
• classify
• what tags
• bridge tags metadata: short-description: Extract taxonomy tags for graph traversal

---

Taxonomy

Extract Federated Taxonomy tags from text for memory storage and multi-hop graph traversal.

Quick Start

# Extract tags from text
./run.sh --text "Error handling in the authentication module" --collection operational

# Extract from file
./run.sh --file document.txt --collection lore

# Just get bridge tags (for graph traversal)
./run.sh --text "..." --bridges-only

Output

{
  "bridge_tags": ["Resilience", "Loyalty"],
  "collection_tags": {"function": "Fix", "domain": "Middleware"},
  "confidence": 0.8,
  "worth_remembering": true
}

Bridge Tags (Tier 0)

Shared across all collections - enable cross-collection queries:

| Tag | Indicates | |-----|-----------| | Precision | Methodical, optimized, algorithmic | | Resilience | Fault tolerance, error handling, robustness | | Fragility | Technical debt, brittleness, single point of failure | | Corruption | Bugs, data corruption, silent failures | | Loyalty | Security, compliance, auth, encryption | | Stealth | Hidden, evasion, infiltration |

Collection Types

• lore - Narrative/story content (HLT vocabulary)
• operational - Code/technical lessons (Operational vocabulary)
• sparta - Security content (ATT&CK/D3FEND vocabulary)

Use Cases

1. Before storing to memory: Get tags to enable graph traversal
2. Filtering what to remember: Check worth_remembering field
3. Cross-collection queries: Use bridge_tags for multi-hop search

Composing with /learn

# /learn can call /taxonomy for tag extraction
tags=$(./run.sh --text "$content" --collection operational --json)
bridge_tags=$(echo "$tags" | jq -r '.bridge_tags | join(",")')