README

Test Runs: Analysis Agent Sequence

Purpose: Test the effectiveness of the slash-command sequence (/discover → /plan_analysis → /review) on real analysis outlines.

How to use:

• Each subfolder = one test run (e.g. 2026-01-20_plan_level_unit_economics/).
• Step 0: Run /prepare_analysis_input on the raw outline (PDF, doc): convert to markdown + Dataset cross-reference (baseline) vs data_platform/ and schema. Output: 00_input_proposal.md.
• 00_input_proposal.md = structured outline + Dataset cross-reference (baseline).
• 01_discover_output.md, 02_plan_analysis_gap_output.md, 03_review_output.md = outputs of the three commands.
• 04_test_protocol.md = how to run, effectiveness criteria, and how to re-run with a PDF.

To test with a PDF: Add the PDF to the run folder, extract/paste its text into 00_input_proposal.md, then re-run the three commands (see 04_test_protocol.md).

To evaluate: Use the scoring table in 04_test_protocol.md §3 (Effectiveness criteria).