Content
38%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 has a legitimate core workflow (NOS three-domain evaluation with script-based formatting), but it is buried in generic template boilerplate, its evaluation criteria and both scripts are delegated to files that are absent from the bundle, and no concrete example of the aggregated JSON or scoring rubric appears inline. As shipped, the skill cannot be executed as documented.
Suggestions
Ship the referenced bundle files (`references/nos_criteria_prompts.md`, `scripts/extract_pdf.py`, `scripts/format_nos_table.py`) or remove all references to them and inline the NOS criteria for the Selection, Comparability, and Exposure domains.
Cut the template filler — Key Features, Example Usage ("See `## Usage` above"), Recommended Workflow, Required Inputs, Output Contract, and Validation and Safety Rules add no NOS-specific information; consolidate the real workflow (Usage + Detailed Workflow + Helper Scripts) into one section.
Add a concrete example of the aggregated JSON passed to `python scripts/format_nos_table.py '<json_string>'` and show the expected summary-table output, replacing the placeholder and the environment-specific `cd "20260316/..."` path.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Noticeably verbose with several padded template sections: "Key Features" restates the description verbatim, "Example Usage" says only "See `## Usage` above for related details", and "Use this skill when the request matches its documented task boundary" is circular filler. Generic boilerplate (Recommended Workflow, Output Contract, Validation and Safety Rules, Failure Handling, Required Inputs) adds tokens without NOS-specific knowledge. It is template padding rather than explanation of known concepts, so it fits anchor 2 rather than anchor 1's 'extensively explains concepts Claude already knows'. | 2 / 5 |
Actionability | Some concrete guidance exists — real commands (`python scripts/extract_pdf.py --help`, `python scripts/format_nos_table.py '<json_string>'`) and a four-step usage sequence — but key details are missing: the actual NOS criteria are delegated to `references/nos_criteria_prompts.md` which does not exist in the bundle, `'<json_string>'` is a placeholder with no example JSON, and the `cd "20260316/scientific-skills/..."` path is environment-specific. This matches anchor 3 ('some concrete guidance but incomplete ... missing key details'), not anchor 4 whose gaps would be minor. | 3 / 5 |
Workflow Clarity | A clear sequence is present (extract metadata → evaluate Selection/Comparability/Exposure → synthesize JSON → format table), but checkpoints are missing or implicit between steps: Steps 1-3 say only "Evaluate the domain" with criteria in a nonexistent file, and the validation text ("Run a final validation pass for completeness, consistency, and safety") is generic template language rather than a real checkpoint tied to the workflow. This fits anchor 3 rather than anchor 4's explicit per-step verify steps. | 3 / 5 |
Progressive Disclosure | The body signals one-level-deep references (`references/nos_criteria_prompts.md`, `scripts/extract_pdf.py`, `scripts/format_nos_table.py`), but the bundle contains no references/, scripts/, or assets/ directories — every referenced path is dangling, so the promised navigation leads nowhere. Structure exists in name only, and the inline template sections are content that belongs in separate files or nowhere. Scored against the actual (empty) bundle structure, this is a broken reference graph, matching anchor 2 rather than anchor 3's 'references present but not clearly signaled'. | 2 / 5 |
Total | 10 / 20 Passed |