Content
88%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 exemplary reference-skill body: dense with non-obvious, verifiable domain facts, fully actionable tool call sequences with parameters, and genuinely strong workflow design — polling loops with terminal states, a pre-run model-checker validation step, and a deliberate safety choice between refactor and forEach for batch mutation. Its only deductions are minor repetition of the reference-validity caveat and the three referenced files being absent from the provided bundle, which leaves progressive disclosure unverified beyond the document itself.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Efficient and information-dense: nearly every line carries non-obvious domain facts ("The reference is only valid until the next console interaction", "refactor ... pops a modal confirmation dialog", "preview is ... cut at 200 characters") and it assumes competence — no padding explaining what MPS or an IDE is. Not a 5: the reference-validity caveat is repeated across items #4, #5, and #8, and several parenthetical asides in "Ways this skill gets used" could be tightened or pushed to a reference file. | 4 / 5 |
Actionability | Fully executable guidance throughout: exact MCP tool names with parameters and formats ("mps_mcp_get_console_history(includeResponses = true, limit = 2)", "mps_mcp_print_node(nodeReference = ..., format = \"PLAIN TEXT\")"), runnable surface-syntax examples ("#instances(Ingredient).select({~it => it.name; })"), and a decision table mapping value shapes to printed output and response nodes. The 'shape the query for what you need back' section gives copy-ready patterns for both values and node references, plus an explicit anti-pattern. | 5 / 5 |
Workflow Clarity | Multi-step processes are clearly sequenced with explicit validation and feedback loops: the output-reading procedure is a numbered poll loop with terminal conditions ("finished when the last entry is a kind:response", "still running ... or it printed nothing", "call again a few times, then stop waiting") and a previewComplete check before trusting or re-printing; way #8 provides pre-run checker validation; and the refactor-vs-forEach directive is an explicit confirmation checkpoint for destructive batch mutation ("when the user asks you to create and run it yourself ... prefer forEach — refactor would ... stall on the dialog"). This matches the anchor of clear sequence with explicit validation steps and error-recovery loops. | 5 / 5 |
Progressive Disclosure | Good structure: a dedicated Reference Index clearly signals three one-level-deep references with accurate content summaries ("references/console-languages.md for the full console command-language reference", "references/smodel-query.md for the ... with statement", "references/mcp-insertion.md for mps_mcp_insert_console_command_from_json"), plus inline pointers in the relevant sections and a Related Skills map. Not a 5: those three referenced files are not present in the provided bundle, so navigation cannot be verified end-to-end, and the body carries more operational detail (the full printing/value-shape table and the #4–#8 MCP caveats) than a pure overview would. | 4 / 5 |
Total | 18 / 20 Passed |