When your Agentforce agent misbehaves in production, this skill helps you figure out why. It queries session trace data from Data Cloud's STDM layer, finds patterns in failures or regressions, then lets you reproduce the exact conversation in preview mode and edit the agent file to fix it. The workflow is observe, reproduce, improve. If your org doesn't have STDM enabled yet, it falls back to running test suites and analyzing local traces instead. One nice detail: it handles the annoying double-nested path bug that sf project retrieve creates and auto-resolves agent names between MasterLabel and DeveloperName formats. Aimed at the moment when someone reports "the agent stopped working" and you need to dig into what actually happened.
npx skills add https://github.com/salesforceairesearch/agentforce-adlc --skill observing-agentforce