Content
55%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is comprehensive and clinically detailed with a sensible bundle structure (real references, templates, and most scripts), but it suffers from repetition of the mandatory-first-page requirement, a fabricated scripts list, and a missing concrete build→validate→retry workflow. It reads as a thorough specification rather than lean, executable guidance.
Suggestions
Remove the four-fold repetition of the mandatory page-1 executive-summary rule; state it once in 'Document Structure' and reference it from 'Output Format' and 'Best Practices'.
Fix the 'Scripts' section: delete the four nonexistent files (create_waterfall_plot.py, create_forest_plot.py, calculate_statistics.py, grade_evidence.py) and add an example invocation for each real script.
Add a concrete end-to-end workflow with a validation checkpoint, e.g. fill template → run scripts → `python scripts/validate_cds_document.py` → fix and re-validate → compile to PDF.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~510-line body is mostly domain-useful but padded: the page-1 executive-summary structure and LaTeX tcolorbox example are repeated across the 'Document Structure', 'Output Format', and 'Best Practices' sections (the mandatory-first-page requirement is stated four times), which is verbosity even though accurate. | 2 / 3 |
Actionability | It gives concrete LaTeX structure and one real executable command (`python scripts/generate_schematic.py`), but the 'Scripts' section names 6 nonexistent files (create_waterfall_plot.py, create_forest_plot.py, calculate_statistics.py, grade_evidence.py) and never shows how to invoke the real analysis scripts, so guidance is partly fabricated and incomplete rather than copy-paste ready. | 2 / 3 |
Workflow Clarity | A multi-step document-generation flow is implied (executive summary → TOC → detailed sections → figures) and validation is mentioned via `validate_cds_document.py`, but there is no sequenced build/validate/retry workflow with checkpoints — the validation script is only listed, never wired into a validate→fix→revalidate loop, capping clarity at 2. | 2 / 3 |
Progressive Disclosure | Content is appropriately split into real `references/`, `assets/`, and `scripts/` bundles signaled by 'See the references/ directory', but navigation is generic rather than per-file, the templates section lists files correctly while the scripts section lists nonexistent ones, and large inline blocks (full LaTeX example, repeated best-practices) that belong in references sit in the body. | 2 / 3 |
Total | 8 / 12 Passed |