What problem does it solve? Debugging an Agentforce agent session requires manually querying dozens of Data Cloud DMOs and stitching together interactions, steps, LLM generations, and gateway requests by hand. This Skill automates that entire reconstruction, turning a raw session id into a readable end-to-end trace. ## Core Features & Use Cases - 24-DMO fetch waterfall: Pulls STDM and GenAI audit DMOs from the Data Cloud Query REST API in 5 dependency-ordered waves using the sf CLI for authentication. - Hierarchical session tree: Joins Session → Interaction → Step → Generation → GatewayRequest into a single JSON source of truth, with declared-binding and timestamp-window fallback logic for orphan gateway calls. - Rendered markdown summary: Produces an up-to-11-section human-readable report covering transcript, per-turn trace, token counts, trust scores, feedback, and empty-DMO diagnostics. - Session discovery: Finds sessions by time range, agent, channel, outcome, or conversation text when no session id is known, and resolves MessagingSession ids (0Mw…) to Agent Session UUIDs. - Use Case: A user says "trace session 019dface-... in my-org" — the Skill runs fetch → assemble → render and returns a full walkthrough of what the agent did on every turn. ## Quick Start Ask the assistant to trace a specific Agentforce session by providing its session UUID or MessagingSession id along with your sf CLI org alias.
