Skills
Claude Outcomes: Independent Grader Agent in Separate Context Window Drives 10-Point Task Success Improvement — Write Rubrics, Not Prompts
Outcomes (public beta May 6) introduces a dedicated grader agent that evaluates output against your rubric in a completely separate context window, with no knowledge of the agent's reasoning path. This independence — not self-critique — is what drives the improvement. Wisedocs cut document review time 50% using it. Pattern: writer drafts, grader fetches every URL and checks every claim, feedback drives revisions until pass.
↳ Follow the thread