What problem does it solve? When a multi-agent Gorgon workflow fails mid-run, hangs, or exhausts its token budget, finding the root cause requires manually digging through checkpoint databases, agent logs, and output contracts. This Skill automates that forensic analysis and produces an evidence-based diagnosis with actionable fixes. ## Core Features & Use Cases - Failure Diagnosis: Locates the exact failure point in the checkpoint database, examines agent outputs against expected schemas, checks per-agent budget consumption, and classifies the root cause into a defined failure taxonomy (contract violation, budget exhaustion, timeout, hallucination, and more). - Post-Mortem Analysis: Analyzes completed workflow runs to identify wasted tokens, slow stages, and optimization opportunities with estimated savings. - Actionable Reporting: Produces a structured debug report with evidence, fix options ordered by ROI with effort estimates, and a single recommended fix. - Use Case: A workflow fails at step 4 of 7 with a cryptic error. Run the diagnosis to discover the upstream agent returned a partial result missing a required key, then receive three fix options ranging from a 15-minute quick patch to a systemic schema-validation fix. ## Quick Start Ask the agent to diagnose why workflow run run_2026-02-12_001 failed using the checkpoint database at .gorgon/checkpoints.db and the agent logs in .gorgon/runs.
