Content
71%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.
An exceptionally actionable, project-specific reference with explicit lifecycles and copy-paste-ready code and commands. Its weaknesses are redundancy (duplicated sections and repeated fetch descriptions), and monolithic structure — all reference material is inlined in one large file with no progressive disclosure layer.
Suggestions
Deduplicate content: the 'gRPC timeouts' section is repeated verbatim in §4 and §8, and the per-entity fetch descriptions in §2 overlap §4 — keep one authoritative copy.
Split reference-heavy material (the theme field table, proto field gotchas, per-screen key hints) into references/ files linked from a leaner SKILL.md overview.
Embed validation checkpoints directly in the change workflows (e.g. 'run cargo clippy -p openshell-tui after adding a screen') instead of deferring verification to an external CONTRIBUTING.md.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Nearly all content is project-specific (RPC names, proto field gotchas, key bindings) with no padding about concepts Claude already knows, but there is real trimmable fat: the 'gRPC timeouts' section appears verbatim in both §4 and §8, the fetch-per-entity descriptions in §2 are re-stated in §4, and the 16-row theme field table plus per-screen key-hint catalog run long. It could be meaningfully tightened without losing clarity. | 3 / 5 |
Actionability | Fully executable guidance throughout: copy-paste-ready commands ('cargo build -p openshell-tui', 'mise run term:dev', 'helm upgrade --install openshell deploy/helm/openshell --namespace openshell'), literal proto request constructions with field names and '..Default::default()', the tokio::spawn/mpsc event pattern, and step-by-step recipes for adding a screen, gRPC call, or Event variant that cover the common change cases. | 5 / 5 |
Workflow Clarity | Multiple explicit numbered lifecycles (log streaming, gateway switching, settings CRUD, startup, workspace switching) with failure paths ('On failure: status_text shows the error') and y/n confirmation gates on destructive operations, plus a Verification section with concrete commands. Validation is not embedded as checkpoints inside the workflows themselves, and verification defers to a CONTRIBUTING.md that is not part of the bundle, which keeps it below the top anchor. | 4 / 5 |
Progressive Disclosure | Internal structure is strong (nine numbered sections, tables, a module dependency tree), but there are no bundle files at all — the 16-field theme table, proto field reference, per-screen key-hint catalog, and file-by-file architecture map are all inlined in a single ~640-line SKILL.md. Reference material of this volume is what the rubric expects split into clearly signaled, one-level-deep reference files. | 3 / 5 |
Total | 15 / 20 Passed |