esm
Comprehensive toolkit for protein language models including ESM3 (generative multimodal protein design across sequence, structure, and function) and ESM C (efficient protein embeddings and representations). Use this skill when working with protein sequences, structures, or function prediction; designing novel proteins; generating protein embeddings; performing inverse folding; or conducting protein engineering tasks. Supports both local model usage and cloud-based Forge API for scalable inference.
89
86%
Does it follow best practices?
Impact
93%
3.44xAverage score across 3 eval scenarios
Passed
No known issues
Quality
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 a strong skill description that clearly identifies its domain (protein language models), lists specific capabilities and model variants, and includes an explicit 'Use this skill when...' clause with comprehensive trigger scenarios. The description is well-structured, uses third person voice throughout, and provides enough specificity to distinguish it from any other skill in a large collection.
| Dimension | Reasoning | Score |
|---|---|---|
Specificity | Lists multiple specific concrete actions: protein design across sequence/structure/function, generating protein embeddings, inverse folding, protein engineering, and distinguishes two model variants (ESM3 and ESM C) with their specific capabilities. | 3 / 3 |
Completeness | Clearly answers both 'what' (comprehensive toolkit for protein language models with specific capabilities listed) and 'when' (explicit 'Use this skill when...' clause covering multiple trigger scenarios like working with protein sequences, designing novel proteins, generating embeddings, etc.). | 3 / 3 |
Trigger Term Quality | Excellent coverage of natural terms a user would say: 'protein sequences', 'structures', 'function prediction', 'protein design', 'protein embeddings', 'inverse folding', 'protein engineering', 'ESM3', 'ESM C', 'Forge API'. These are the terms domain users would naturally use. | 3 / 3 |
Distinctiveness Conflict Risk | Highly distinctive niche focused on protein language models (ESM3/ESM C) with domain-specific triggers like 'inverse folding', 'protein embeddings', and 'protein engineering' that are unlikely to conflict with other skills. | 3 / 3 |
Total | 12 / 12 Passed |
Implementation
72%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This is a well-structured skill with strong actionability through executable code examples covering all major ESM capabilities. Its main weaknesses are moderate verbosity (explanatory text and 'when to use' sections that Claude doesn't need) and lack of validation/error-handling steps in workflows. The progressive disclosure pattern is excellent with clear references to detailed documentation.
Suggestions
Add validation checkpoints to workflows, e.g., checking that generated sequences are valid amino acid sequences, verifying structure prediction quality scores, and handling API errors with retry logic.
Remove 'When to use' bullet lists and the 'Resources and Documentation' section with blog links—Claude doesn't need these to execute tasks. This would save ~40 lines.
Add error handling examples for common failure modes (e.g., Forge API timeout, invalid sequence input, CUDA out of memory) to improve workflow robustness.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The skill includes some unnecessary framing (e.g., 'When to use' lists that Claude can infer, the 'Overview' section restating the description, the 'Responsible Use' and 'Resources' sections with blog links). However, the code examples are reasonably tight and the model selection guide is genuinely useful. Could be trimmed by ~30%. | 2 / 3 |
Actionability | The skill provides fully executable, copy-paste ready Python code for all major use cases: sequence generation, structure prediction, inverse folding, embeddings, function conditioning, chain-of-thought generation, and batch processing. Installation commands are concrete. Model names and API endpoints are specific. | 3 / 3 |
Workflow Clarity | The chain-of-thought generation section shows a clear multi-step workflow, but there are no validation checkpoints anywhere. For protein design tasks (which are complex, multi-step, and potentially destructive in terms of compute cost), there's no guidance on validating outputs, checking for errors, or handling generation failures. The workflows are sequential but lack feedback loops. | 2 / 3 |
Progressive Disclosure | Excellent progressive disclosure with a clear overview in the main file and well-signaled one-level-deep references to esm3-api.md, esm-c-api.md, forge-api.md, and workflows.md. The references section clearly describes what each file contains, making navigation easy. | 3 / 3 |
Total | 10 / 12 Passed |
Validation
90%Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.
Validation — 10 / 11 Passed
Validation for skill structure
| Criteria | Description | Result |
|---|---|---|
metadata_version | 'metadata.version' is missing | Warning |
Total | 10 / 11 Passed | |
- Repository
- K-Dense-AI/claude-scientific-skills
- Commit
b58ad7e
- 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.