Content
56%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 contains a strong domain core — IRB/EC essential elements, prohibited content rules, structured inputs, and an executable script path — wrapped in a large amount of template boilerplate that inflates token cost without adding guidance. The References section points at files that do not exist, and the main example embeds a non-portable personal path, both of which would confuse an executing agent.
Suggestions
Delete or collapse the boilerplate sections (Key Features, Dependencies, Implementation Details, Risk Assessment, Security Checklist, Evaluation Criteria, Lifecycle Status) and the repeated scope statements; they roughly double the file without adding skill-specific instruction.
Fix the References section to match the actual bundle: remove the nonexistent 'plain_language_guide.pdf' and correct 'template_examples/' to the real file 'references/template_examples.md'.
Replace the hardcoded '/Users/z04030865/.openclaw/...' path in the example with the relative 'scripts/main.py' form already used elsewhere, and add an explicit output-validation step (check generated ad against the Prohibited Content list before returning it).
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is noticeably padded: the scope sentence is repeated verbatim in three sections; 'Key Features', 'Dependencies', 'Example Usage', and 'Implementation Details' open with self-referential filler ('See ## Prerequisites above for related details') that adds no information; and generic boilerplate (Risk Assessment table, Security Checklist, Evaluation Criteria, Lifecycle Status) consumes a large share of the file without teaching anything Claude doesn't know. It escapes score 1 only because the genuinely domain-specific material (IRB/EC essential elements, prohibited content, reading-level requirements) is not something Claude already knows and earns its place. | 2 / 5 |
Actionability | Concrete guidance dominates the useful parts: an explicit input parameter block, a full multi-flag example invocation of scripts/main.py, an output structure specification, and executable verification commands ('python -m py_compile scripts/main.py', 'python scripts/main.py --help'). It falls short of 5 because the primary example hardcodes a personal absolute path ('/Users/z04030865/.openclaw/workspace/skills/...'), which is not copy-paste portable, and the referenced example directory '20260318/scientific-skills/...' does not exist in the bundle. | 4 / 5 |
Workflow Clarity | The Workflow section gives a clear five-step sequence with an explicit validation checkpoint (Quick Check compile step), an early-stop rule for out-of-scope requests, and a fallback path when execution fails or inputs are incomplete ('If scripts/main.py fails, report the failure point... and provide a manual fallback'). Not a 5 because the workflow steps themselves are abstract ('Return a structured result that separates assumptions, deliverables, risks') rather than concrete per-step commands, and there is no validation of the generated output (e.g., checking the ad against the prohibited-content list). | 4 / 5 |
Progressive Disclosure | References are one level deep and clearly signaled, but the file listing does not match the actual bundle: 'plain_language_guide.pdf' and the 'template_examples/' directory do not exist in references/, while the real file 'template_examples.md' is mislabeled. Meanwhile roughly 90 lines of detailed compliance checklists (Essential Elements, Prohibited Content) and generic admin sections are inlined in SKILL.md that could live in reference files. This sits between 'references present but not clearly signaled / content that should be separate is inline' (3) and 'references mostly clear, minor organization gaps' (4) — the phantom references push it to 3. | 3 / 5 |
Total | 13 / 20 Passed |