Operational scorecards
Score task completion, tool discipline, evidence quality, safety, and communication.
Agent Evaluation Field Notes
Practical scorecards for tool-using AI agents.
Compact templates, replay checklists, and RAG guardrail smoke tests for builders who need repeatable agent evaluation workflows.
Trajectory replay
Debug agent runs by replaying decisions, tool calls, evidence, and recovery points.
RAG guardrails
Smoke-test prompt injection, vector poisoning, source grounding, and citation behavior.