rdkit
Cheminformatics toolkit for fine-grained molecular control. SMILES/SDF parsing, descriptors (MW, LogP, TPSA), fingerprints, substructure search, 2D/3D generation, similarity, reactions. For standard workflows with simpler interface, use datamol (wrapper around RDKit). Use rdkit for advanced control, custom sanitization, specialized algorithms.
81
75%
Does it follow best practices?
Impact
93%
1.38xAverage score across 3 eval scenarios
Passed
No known issues
Optimize this skill with Tessl
npx tessl skill review --optimize ./scientific-skills/rdkit/SKILL.mdQuality
Discovery
100%Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.
This is an excellent skill description that packs dense, specific information into a concise format. It lists concrete capabilities with domain-appropriate trigger terms, explicitly differentiates from the related datamol skill, and provides clear guidance on when to select this skill. The only minor improvement would be slightly more explicit 'Use when...' phrasing, but the existing guidance is functionally equivalent and effective.
| Dimension | Reasoning | Score |
|---|---|---|
Specificity | Lists multiple specific concrete actions: SMILES/SDF parsing, descriptors (MW, LogP, TPSA), fingerprints, substructure search, 2D/3D generation, similarity, reactions. These are highly specific cheminformatics operations. | 3 / 3 |
Completeness | Clearly answers 'what' (cheminformatics toolkit with specific capabilities listed) and 'when' (use rdkit for advanced control, custom sanitization, specialized algorithms; contrasted with datamol for simpler workflows). The explicit guidance on when to use this vs. datamol serves as a strong 'Use when' clause. | 3 / 3 |
Trigger Term Quality | Excellent coverage of natural terms a cheminformatics user would use: SMILES, SDF, molecular descriptors, MW, LogP, TPSA, fingerprints, substructure search, similarity, reactions, RDKit. These are the exact terms domain users would mention. | 3 / 3 |
Distinctiveness Conflict Risk | Highly distinctive niche in cheminformatics/molecular science. Explicitly differentiates itself from the related datamol skill by specifying when to use each, which directly reduces conflict risk between the two most likely overlapping skills. | 3 / 3 |
Total | 12 / 12 Passed |
Implementation
50%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This skill is highly actionable with excellent, executable code examples covering RDKit's full API surface. However, it is far too verbose for a skill file—it reads more like comprehensive documentation than a concise skill guide, with extensive content that should be delegated to the referenced bundle files. The lack of validation checkpoints in multi-step workflows and the monolithic structure significantly reduce its effectiveness as a skill.
Suggestions
Reduce the main body to ~100-150 lines covering the most essential patterns (molecule I/O, basic descriptors, fingerprints, substructure search) and move detailed API coverage into the referenced files like api_reference.md and descriptors_reference.md.
Remove explanatory text that Claude already knows (e.g., 'RDKit is a comprehensive cheminformatics library providing Python APIs...', descriptions of what LogP or TPSA are) and keep only the code patterns and critical gotchas.
Add validation checkpoints to multi-step workflows, e.g., check EmbedMolecule return value (-1 = failure), verify reaction products are non-empty before sanitization, and validate conformer generation succeeded.
Eliminate duplication such as the Lipinski rule-of-five appearing both in the descriptors section and the common workflows section.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The skill is extremely verbose at ~500+ lines, explaining many concepts Claude already knows (what SMILES is, what fingerprints are, basic Python patterns). Sections like 'Common Pitfalls' restate what's already covered. The Lipinski rule-of-five is explained twice. Much of this is standard RDKit documentation that doesn't need to be in a skill file. | 1 / 3 |
Actionability | The code examples are concrete, executable, and copy-paste ready throughout. Every section includes working Python code with proper imports, and the common workflows section provides complete, runnable functions. | 3 / 3 |
Workflow Clarity | Individual code snippets are clear, but multi-step workflows lack explicit validation checkpoints and sequencing. For example, the 3D coordinate generation workflow doesn't validate embedding success (EmbedMolecule returns -1 on failure), and the reaction workflow doesn't check for empty product sets. The common workflows section provides functions but no step-by-step process with verification. | 2 / 3 |
Progressive Disclosure | The skill references external files (references/, scripts/) at the bottom, which is good, but the main body is a monolithic wall of content that should have much more pushed into those reference files. The 500+ line body contains API reference-level detail (e.g., all fingerprint types, all descriptor functions) that belongs in the referenced api_reference.md and descriptors_reference.md files. | 2 / 3 |
Total | 8 / 12 Passed |
Validation
81%Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.
Validation — 9 / 11 Passed
Validation for skill structure
| Criteria | Description | Result |
|---|---|---|
skill_md_line_count | SKILL.md is long (779 lines); consider splitting into references/ and linking | Warning |
metadata_version | 'metadata.version' is missing | Warning |
Total | 9 / 11 Passed | |
- Repository
- K-Dense-AI/claude-scientific-skills
- Commit
cbcae7b
- Reviewed
Table of Contents
Is this your skill?
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.