# Run the fictional Workspace evaluation
1. Download brief.md, milestones.csv, and meeting-notes.md. Supply only these
   three initial inputs to the candidate. Copy the prompt from the template.
2. Inspect its saved report. Use expected-result.md and scorecard.csv yourself;
   do not supply the answer reference as initial context.
3. Supply correction.md as a second turn. Inspect the revised saved file and
   change note, including dates and owner statements elsewhere in the report.
4. Record pass, revise, or not observed in the scorecard, with evidence.
5. Use the same files and correction in each candidate. Record its actual plan,
   enabled tools, manual steps, review effort, and observed usage.
All names and records are fictional. The expected result is authored, not a
product benchmark. Opening Workspace does not upload these files or run a task.