Content
81%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 highly actionable, well-validated reasoning guide: every workflow ends in a verification or sanity-check step and every script comes with executable example commands. The weaknesses are redundancy — textbook restatements, the DoU formula repeated four times, and tool guidance duplicated across sections 4 and 6 — plus an undocumented bundle script.
Suggestions
Consolidate the tool-vs-fallback guidance: sections 4 and 6 repeat the same "Preferred: use X tool. Fallback: run X.py" lines for crystal_validator, degrees_of_unsaturation, molecular_formula, and smiles_verifier — state each once and cross-reference.
Trim textbook basics Claude already knows (enantiomer/diastereomer/meso definitions, CIP ranking by atomic number, the n+1 splitting rule, standard NMR/IR shift ranges) and state the DoU formula once.
Document or remove the undocumented bundle script `molecular_complexity.py`; consider moving the named-reaction decision tree and retrosynthesis walkthroughs to a references/ file to keep SKILL.md a lean overview.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly dense, high-value heuristics (decision-tree table, base-size/Zaitsev rules, retrosynthetic disconnection heuristics), but several portions restate textbook basics Claude already knows (enantiomer/diastereomer definitions, CIP atomic-number ranking, the n+1 rule, standard shift ranges), the DoU formula appears four times, and the "Preferred tool / Fallback script" guidance is duplicated between sections 4 and 6 — fits "mostly efficient but could be tightened" rather than 4. | 3 / 5 |
Actionability | Copy-paste-ready commands with real example values for every bundled script (e.g. `python crystal_validator.py --a 5.43 --b 5.43 --c 5.43 --Z 8 --MW 28.085 --density 2.329`), named MCP tools with their parameters (`OPSIN_name_to_structure` param `name`, RXN async polling params), and concrete constraint flags — fully executable guidance covering the common cases. | 5 / 5 |
Workflow Clarity | The 5-step product-prediction process ends in an explicit sanity-check step (valences, mass balance, DoU); the spectroscopy strategy ends with "Verify: does your proposed structure predict every observed signal?"; and SMILES design has a feedback loop ("If any constraint FAILs, revise and re-run until all pass") — clear sequences with explicit validation and error-recovery loops. | 5 / 5 |
Progressive Disclosure | Well-sectioned body, and every referenced script exists in the bundle with clear paths and usage examples. However, all reference-style material (script documentation, named-reaction table, retrosynthesis walkthroughs) is inlined in one ~365-line SKILL.md with no separate reference files, and one bundled script (`molecular_complexity.py`) is never mentioned — good structure with minor organization gaps rather than a clean overview pointing to one-level-deep references. | 4 / 5 |
Total | 17 / 20 Passed |