Content
67%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 delivers a well-sequenced, mostly executable routing workflow with genuine validation checkpoints and clear failure modes. Its weaknesses are repetition — the scope-isolation rule is stated four times — and placeholder-laden core call signatures that leave the central staffing operation not fully copy-paste ready.
Suggestions
State the owner-scope/no-Hub/no-legacy-fallback rule once (e.g., in Hard rules) and reference it elsewhere, cutting roughly 15-20% of the body.
Replace the ellipsis placeholders in the workforce call sequence with a minimal concrete example of a WorkOrder/selection payload, or point to a reference file containing one.
Show the actual 'cloud ... --project .' command line referenced in section 1.5, since the first-contact contract depends on it.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The owner-scope/no-Hub/no-legacy-cargo rule is restated four times ("does **not** search the public marketplace" in §0, "never fall back to the legacy cargo search path" in §1, "do not silently query public Hub or legacy cargo search" in §3, "Never report public marketplace agents" in §5) and receipt preservation is repeated across §3 and §4, so the body could be meaningfully tightened rather than just trimmed at the edges. | 3 / 5 |
Actionability | The runner-resolution and auth snippets are copy-paste ready and the Workforce call sequence names concrete typed parameters, but §3 uses ellipsis placeholders ("workOrder=..., selection=...") and §1.5 references a "cloud ... --project ." call whose actual command line is never shown, leaving minor gaps. | 4 / 5 |
Workflow Clarity | Sections 0-5 give a clear sequence (resolve runner → sign in → search → validate → prepare → validate receipt) with explicit checkpoints and error paths ("report source_not_supported", installer fallback, headless env var), but there is no fix-and-retry feedback loop for validation failures, so it stops short of the 5 anchor. | 4 / 5 |
Progressive Disclosure | The skill is a single self-contained file with no bundle files (no references/, scripts/, or assets/ exist and none are cited) and clear numbered sections; at ~112 lines with all content inline, structure is good but it exceeds the under-50-line simple-skill case and has minor organization gaps (e.g., the runner script could live in a scripts/ file). | 4 / 5 |
Total | 15 / 20 Passed |