Content
75%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.
A well-built skill body: consistent, executable workflow examples with a clear submit-wait-fetch pattern, real error-handling guidance, and a verified one-level-deep reference bundle. The main improvements are removing the redundant 'Common Patterns' section, fixing non-executable placeholders (missing stjames install, [x, y, z] pocket), and moving more API detail into the existing reference files.
Suggestions
Add `stjames` to the installation command (e.g., `uv pip install rowan-python stjames`) and replace placeholder values in Pattern 2 (`pocket = {"center": [x, y, z]}` and undefined `compound_library`) with concrete values so every example is copy-paste executable.
Cut the 'Common Patterns' section or reduce it to non-duplicative recipes — Pattern 1 and Pattern 3 restate the Core Workflows examples almost verbatim, padding the token budget without new information.
Move the Workflow Management code and RDKit-native function lists into references/api_reference.md and references/rdkit_native.md, keeping only a pointer in SKILL.md, and add an explicit status check (e.g., `if workflow.status == "failed"`) before accessing `workflow.data` in the main workflow examples.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dominated by executable code with almost no explanation of concepts Claude already knows, fitting the 'efficient; minor instances of over-explanation' anchor. The 'Common Patterns' section (Pattern 1 repeats the pKa workflow, Pattern 3 repeats the conformer search) and the marketing-flavored 'Why Rowan' bullets are the trimmable padding that keeps it below score 5. | 4 / 5 |
Actionability | Core workflows give copy-paste-ready code with the full submit → wait_for_result → fetch_latest → access-data cycle, plus install and auth steps, matching score 4 ('minor gaps'). It is not score 5 because some examples are not directly executable: `stjames` is imported throughout but missing from the install command, and Pattern 2 uses placeholders like `pocket = {"center": [x, y, z]}` and an undefined `compound_library`. | 4 / 5 |
Workflow Clarity | Every workflow follows the same explicit submit → wait → fetch → read-data sequence, and the Error Handling section adds status checks ('if workflow.status == "completed"'), a timeout, and exception handling, giving most checkpoints. It falls short of score 5 because the main workflow examples access `workflow.data` without an explicit status/failed check — validation lives only in a separate section — which is the 'minor validation gaps' pattern of score 4, not the absent-validation pattern of score 3. | 4 / 5 |
Progressive Disclosure | The six referenced files (api_reference.md, workflow_types.md, rdkit_native.md, molecule_handling.md, proteins_and_organization.md, results_interpretation.md) all exist, are described inline, and contain no nested pointers, so references are one level deep and well signaled. Score 4 rather than 5 because the body itself carries substantial API detail inline (full Workflow Management code, RDKit function lists, Computational Methods) that belongs in the reference files under the 'content appropriately split' anchor. | 4 / 5 |
Total | 16 / 20 Passed |