Content
96%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 a dense, high-signal operating manual: precise tool/parameter documentation, a validated lifecycle with explicit checkpoints and error-recovery guidance, and no wasted tokens. The only structural weakness is that all reference material (settings enums, credential-type table) is inlined in SKILL.md rather than split into a references file.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Every token carries API-specific knowledge Claude cannot know — defaults ('removeOrphanedSourcesWhenDisappeared (default true)'), enum values, and gotchas ('Deleting a credential still referenced by a datasource makes its runs fail'). It never explains concepts Claude already knows (what S3 is, how buckets work), matching the 'every token earns its place' anchor. | 5 / 5 |
Actionability | Fully executable for an MCP-tool skill: exact tool names, parameter names with types and defaults, return shapes ({credentialId, ..., batchIds?}, {checks, errors}), a concrete glob example (["images/**/*.jpg", "batch-*/**"]), and a credential-type table enumerating secret fields. The fast path covers the common case end-to-end, so the score-4 'minor gaps' anchor does not apply. | 5 / 5 |
Workflow Clarity | The numbered Create → Validate → Trigger → Poll lifecycle has an explicit validation checkpoint ('Run it to self-diagnose a bad credential/bucket before triggering'), gating ('starts a mirror run only if validation passes'), and error recovery ('status: "partial" with the ids so you can clean up'). Validation is present, so the destructive/batch cap at 3 does not apply, and the sequence matches the explicit-validation-and-feedback-loops anchor. | 5 / 5 |
Progressive Disclosure | Good structure with a fast path up front, clearly labeled sections, and one-level-deep, clearly signaled links (skill://roboflow/roboflow-data-management/SKILL.md). However, everything is inlined (~135 lines, no bundle files): the mirror_configs settings enumeration and credential-type table are reference material living in the overview rather than split out, which is exactly the 'minor organization gaps' of the score-4 anchor and short of the score-5 'content appropriately split' anchor. | 4 / 5 |
Total | 19 / 20 Passed |