Content
77%Weight 40%Scale 1-3Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is highly actionable with a clear, validated workflow and well-signaled one-level-deep bundle references. Its main weakness is verbosity from heavy cross-section repetition and large chunks of reference-level detail kept inline rather than delegated to the reference files.
Suggestions
Consolidate the repeated side-scroll rules (scenery-only background, base must not bake runtime objects, at-most-9 candidates, reference mockup is not a deliverable) into one authoritative section and reference it, instead of restating them in Overview, Workflow, Stage Segments, Playable Stage Reference Rules, Expected Deliverables, and Validation.
Move the long Side-Scroll Stage Segments / Shared Platform Library / Playable Stage Reference Rules detail into a dedicated reference file (e.g. references/side-scroll-contract.md) and keep SKILL.md as a concise overview pointing to it, mirroring the existing layered-map and prop-pack contracts.
Tighten the Parameter Contract and Genre Routing tables by removing default-axis lists already stated in the Map Modes section to reduce duplicated inline detail.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body avoids explaining concepts Claude already knows, but it restates the same rules many times — the scenery-only-background, base-must-not-bake-runtime-objects, at-most-9-candidates, and reference-mockup-is-not-final rules each recur across Overview, Workflow, Side-Scroll Stage Segments, Playable Stage Reference Rules, Expected Deliverables, and Validation, so it could be tightened considerably. | 2 / 3 |
Actionability | It provides an executable bash invocation for scripts/extract_terrain_tiles.py (arguments verified against the script), concrete asset path conventions (assets/map/<name>-base.png), full parameter enums, and per-mode deliverable lists that are copy-paste ready. | 3 / 3 |
Workflow Clarity | A numbered Workflow (steps 1–5) ends with a dedicated Validate-and-preview step, reinforced by a Validation checklist and the Post-Reference Object Production Gate, with explicit feedback loops (regenerate/reject mismatched or object-baked bases) for batch image-generation operations. | 3 / 3 |
Progressive Disclosure | The three references and three scripts all exist, are one level deep, and are clearly signaled via markdown links, but the 460-line body keeps large reference-level detail inline — notably the side-scroll rules repeated across several sections — so content that should be split out remains in SKILL.md. | 2 / 3 |
Total | 10 / 12 Passed |