Content
100%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 an exemplary lean instruction skill: concrete tool and REST details, a copy-paste request example, explicit data-handling rules (unknown vs zero, truncation, provenance), error/retry policy, and injection-defense guidance, all in well-organized sections with zero padding. No weaknesses surfaced against any dimension.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and assumes competence: it never explains what CPI or yields are, and every line carries operational facts (dataset table, "limit defaults to 30 and accepts 1..365", response shape, routes, auth header, error policy). This matches anchor 5 — every token earns its place; nothing matches the padding of anchors 3–4. | 5 / 5 |
Actionability | Guidance is fully concrete: the tool name `get_macro_history` with server URL, a dataset table with required parameters, a copy-paste request example ("{\"dataset\":\"world-cpi\",\"country\":\"Germany\",\"limit\":12}"), exact REST routes, and the precise auth header "X-WorldMonitor-Key". Per the rubric's code_vs_instruction note, an instruction-only skill with guidance this specific is not penalized for lacking code; it matches anchor 5's 'specific examples cover the common cases'. | 5 / 5 |
Workflow Clarity | The flow is sequenced with explicit checkpoints: "Read `unavailable` first", "`truncated: true` means the response is not full history", "request one country at a time and align observation dates and definitions before comparing", and the worked example ends with comparing on common dates, explaining methodology differences, and citing source labels. Retry guidance ("Respect billing denials, rate limits and `Retry-After`") covers error recovery; the destructive/batch cap does not apply to read-only retrieval, and the under-50-line simple-skill exception applies. | 5 / 5 |
Progressive Disclosure | The skill is under 50 lines with no bundle files and no need for external references, and the body is cleanly sectioned (Connected plugin, REST and authentication, Errors, Content safety) with no dangling or buried references. Per the judging guidelines, this matches the simple-skill case where well-organized sections alone score 5. | 5 / 5 |
Total | 20 / 20 Passed |