| name | cxas-protocol-two-phase-ingestion |
| description | A two-phase protocol for extracting structure and generating transcripts from customer artifacts. |
Two-Phase Ingestion Protocol
This protocol defines a strict two-phase process for analyzing customer
artifacts to extract system structure and generate high-fidelity conversation
transcripts.
Phase 1: Structural Digestion
- Goal: Extract raw data from artifacts into a digestible form (e.g., YAML
or structured Markdown).
- Scope: Map the exact Taxonomy, Intents, Tools, Connections, state
transitions, and node logic.
- Rule: DO NOT infer transcripts during this phase. Focus purely on
structural extraction.
- Exception: If explicit transcripts are already present in the raw data
(e.g., inside Cyara XML test data), they can be extracted directly.
Phase 2: Transcript Generation
- Goal: Generate full natural conversation transcripts based on the
digested structural data.
- Mechanism: Use LLM reasoning to read the structural data and generate
transcripts.
- Rule: Must use the
./append_turn.py script (located in the parent skill directory)
to build transcripts turn-by-turn to ensure schema compliance.