inspect-agent-run-with-prefactor-cli

To perform root-cause analysis on a Prefactor agent run, provide the agent repository path, agent ID, and agent instance ID. The Skill correlates span data with local source code to build a structured findings report.

Agent runs producing unexpected results or high costs are diagnosed by inspecting Prefactor span data. The Skill correlates spans with repository context to identify triggers and pinpoint where behavior diverged from expected outcomes.

To generate a post-mortem findings report for agent spans, you need access to the agent's local repository, Prefactor CLI, and the specific span data. This context allows correlation of traces with source code locations.

Span analysis identifies the root causes of agent-run issues but does not prescribe fixes. It supports safe post-mortem analysis by focusing on evidence-backed observations, timelines, and source locations rather than remediation.

Prefactor CLI tracing works with local source code references by correlating span data with the agent's repository context. This process maps timeline events to specific prompts and source locations to identify behavioral divergence.

Span analysis for Prefactor agent instances is the process of inspecting execution traces to understand run issues. It builds evidence-backed findings reports by correlating timeline events with local code references to determine root causes.