Content
46%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 core sections (When to Use, Example Usage, Implementation Details, Reference pointer) are well-constructed with executable commands and a real one-level-deep bundle, but the body is dragged down by generic template boilerplate and documentation that misrepresents the script's actual interface (evidence type and fields flag unsupported, search type missing). Trimming the filler and aligning the docs with the script would lift most dimensions.
Suggestions
Remove the generic template sections (When Not to Use, Required Inputs, Deterministic Output Rules, Output Contract, Validation and Safety Rules, Failure Handling, Completion Checklist, Scope Reminder) — they restate practices Claude already knows and roughly double the token cost without skill-specific value.
Align documented usage with the actual script: either implement `--type evidence` and `--fields` or drop them from the docs, and document the supported `search` query type that the script actually provides.
Replace the fabricated "Quick Validation" expected output (which `--help` does not produce) with the real help text, and rewrite Recommended Workflow as the concrete query steps (choose entity type → run script → parse JSON output) instead of template language.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Roughly half the body is generic template filler — "Deterministic Output Rules", "Output Contract", "Validation and Safety Rules", "Completion Checklist", "Scope Reminder", "Required Inputs" — that restates practices Claude already knows ("Keep the output safe, reproducible, and within the documented scope"). Several padded sections with no skill-specific content put this noticeably below the midpoint; the core sections (When to Use, Example Usage, Implementation Details) are efficient, keeping it above a 1. | 2 / 5 |
Actionability | The target-query bash commands are copy-paste ready and inputs/outputs are concretely specified, but key details are wrong or missing: the documented `--type evidence` is not supported by the script (choices are target/disease/search), the `--fields` option is unimplemented and hedged with "if supported by your script/API wrapper", and the script's actual `search` type is undocumented. This mismatch makes the guidance incomplete despite being executable in part. | 3 / 5 |
Workflow Clarity | The "Recommended Workflow" is a generic four-step template ("Select the documented execution path", "Produce the expected output using the documented file format, schema, or narrative structure") not tied to the actual query process, and "Quick Validation" shows an expected output ("Result file: open_targets_db_result.md") that `--help` does not produce. A concrete sequence exists via the example commands, but checkpoints and the real workflow are implicit or fabricated. | 3 / 5 |
Progressive Disclosure | The bundle structure is genuinely two-tier: SKILL.md serves as overview with a clearly signaled, one-level-deep reference ("See `references/api_reference.md` for API details, available fields, and entity schemas") and the referenced file exists with appropriate API detail. Minor gaps only: the single reference is fine, but the boilerplate sections clutter navigation and some API detail (unsupported fields list) leaks into the body. | 4 / 5 |
Total | 12 / 20 Passed |