Content
50%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is well structured with useful intent-mapping tables and a prompt-expansion formula, but every dimension lands at the middle anchor: it defines concepts Claude already knows, its example command is syntactically broken, the workflow lacks validation/error-handling checkpoints, and it references scripts that are missing from the bundle.
Suggestions
Fix the bash example by adding '\' line continuations (or putting it on one line) and show the expected output including the request_id, so the invocation is copy-paste executable.
Ship the referenced scripts (scripts/generate-film.sh, core/platform/check-result.sh) in the bundle or remove/correct the references — currently the core workflow points to nonexistent files.
Replace the glossary-style definitions of standard camera/lens terms with the model-specific syntax they should produce, and give one worked example of a raw prompt expanded via the Technical Infusion formula.
Make Step 3 concrete: include the actual check-result.sh polling command, expected completion response, and what to do on failure or timeout.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient tables and formulas, but it defines standard cinematography terms Claude already knows ('Dolly In/Out: Physical camera movement on a track...', 'Shallow DOF: Background blur (Bokeh)') and the 'Core Competencies' section duplicates content from the Intent Mapping Table and Prompt Optimization Protocol. | 3 / 5 |
Actionability | There is a concrete bash invocation with flags, but the example command is broken as written (multi-line invocation missing '\' continuations), references scripts absent from the bundle, and provides no example of an expanded prompt or expected script output — matching 'some concrete guidance but incomplete'. | 3 / 5 |
Workflow Clarity | The three-step protocol (define brief → invoke script → poll result) has a clear sequence, but Step 1 is vague ('Provide the agent with a subject and a Director's Intent'), and Step 3 gives no polling command, interval, or failure handling, leaving checkpoints implicit. | 3 / 5 |
Progressive Disclosure | Sections are well organized with clear headers, but the body references 'scripts/generate-film.sh' and 'core/platform/check-result.sh' — neither file exists in the bundle (no scripts/ directory), so the core workflow is unverifiable and navigation is broken. | 3 / 5 |
Total | 12 / 20 Passed |