$npx -y skills add GoogleCloudPlatform/cxas-scrapi --skill cxas-protocol-two-phase-ingestionA two-phase protocol for extracting structure and generating transcripts from customer artifacts.
| 1 | # Two-Phase Ingestion Protocol |
| 2 | |
| 3 | This protocol defines a strict two-phase process for analyzing customer |
| 4 | artifacts to extract system structure and generate high-fidelity conversation |
| 5 | transcripts. |
| 6 | |
| 7 | ## Phase 1: Structural Digestion |
| 8 | |
| 9 | - **Goal**: Extract raw data from artifacts into a digestible form (e.g., YAML |
| 10 | or structured Markdown). |
| 11 | - **Scope**: Map the exact Taxonomy, Intents, Tools, Connections, state |
| 12 | transitions, and node logic. |
| 13 | - **Rule**: DO NOT infer transcripts during this phase. Focus purely on |
| 14 | structural extraction. |
| 15 | - **Exception**: If explicit transcripts are already present in the raw data |
| 16 | (e.g., inside Cyara XML test data), they can be extracted directly. |
| 17 | |
| 18 | ## Phase 2: Transcript Generation |
| 19 | |
| 20 | - **Goal**: Generate full natural conversation transcripts based on the |
| 21 | digested structural data. |
| 22 | - **Mechanism**: Use LLM reasoning to read the structural data and generate |
| 23 | transcripts. |
| 24 | - **Rule**: Must use the `./append_turn.py` script (located in the parent skill directory) |
| 25 | to build transcripts turn-by-turn to ensure schema compliance. |