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 skill is genuinely executable — concrete commands, a parameter table, CSV format, sample output, and explicit error-handling paths — but it is padded with four overlapping generic process sections, and its documented script behavior is stale: the 'known NameError bug' is already fixed while the actual --institution flag goes undocumented, producing misleading guidance.
Suggestions
Update the script documentation to match the bundled code: remove the stale NameError-bug warning and Claude Direct Path workaround for check_institutional_conflict(), and document the existing --institution flag in the Parameters table.
Consolidate the four overlapping generic sections ('Workflow', 'Stress-Case Rules', 'Output Requirements', 'Response Template') into one short checklist, cutting ~40 lines of boilerplate.
Rewrite the Workflow section as the actual operational sequence (validate inputs -> run script with --publications when available -> interpret report -> report assumptions and demo-data fallback) with explicit checkpoints instead of generic meta-steps.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The core usage content (commands, parameters, CSV format, example output) is tight, but roughly 40 of 147 lines are generic process scaffolding ('Workflow', 'Stress-Case Rules', 'Output Requirements', 'Response Template') with heavy mutual overlap — several padded sections that Claude does not need spelled out, matching the level-2 anchor. | 2 / 5 |
Actionability | Commands are copy-paste ready, but key details are wrong or missing: the claimed NameError bug in check_institutional_conflict() is stale (the bundled script is fixed and exposes an undocumented --institution flag), so the skill redirects Claude away from a working script path to a manual workaround, and the parameter table omits a real CLI option. | 3 / 5 |
Workflow Clarity | A numbered sequence exists, but the 'Workflow' section is generic meta-process ('Confirm the user objective, required inputs, and non-negotiable constraints') rather than the operational run-check-interpret sequence; real validation checkpoints (compile check, error handling, fallbacks) are spread across sections and left implicit. | 3 / 5 |
Progressive Disclosure | The body is well-sectioned with clear headers, the sole bundle reference (scripts/main.py) is real and correctly linked, and no bulky reference material is inlined; the main gap is boilerplate sections that belong (if anywhere) in a condensed checklist rather than four overlapping blocks. | 4 / 5 |
Total | 12 / 20 Passed |