CtrlK
BlogDocsLog inGet started
Tessl Logo

tooluniverse-protein-therapeutic-design

AI-guided de novo protein design — RFdiffusion backbone generation, ProteinMPNN sequence design, structure validation (pLDDT, pTM, MPNN scores). Use for designing therapeutic protein binders, novel scaffolds, enzyme variants, and miniprotein/protein-interface design before experimental validation.

69

Quality

83%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

SecuritybySnyk

Low

Low-risk findings worth noting

The canonical home for this skill is tooluniverse-protein-therapeutic-design in mims-harvard/ToolUniverse

SKILL.md
Quality
Evals
Security

Quality

Content

75%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

A well-structured, tool-centric skill body with concrete parameters, a clear phased workflow, and a validation/checklist backbone. The main gaps are generic exhortation sections that dilute conciseness, missing explicit retry feedback loops for failed designs, and reference files that are named but not bundled for verification.

Suggestions

Trim or fold the generic 'LOOK UP, DON'T GUESS' / 'COMPUTE, DON'T DESCRIBE' sections into the workflow phases so every line teaches protein-design-specific behavior rather than general agent conduct.

Add an explicit feedback loop after Phase 4 validation (e.g. 'if pLDDT < 70 or backbone RMSD high, regenerate backbones or re-run ProteinMPNN with adjusted temperature') to close the batch-operation retry gap.

Signpost reference files inline at the relevant phase (e.g. 'Phase 2: see DESIGN_PROCEDURES.md for sampling parameters') and ensure the referenced .md files are actually bundled so links resolve.

DimensionReasoningScore

Conciseness

Mostly efficient with tool/parameter tables and checklists, but the generic 'LOOK UP, DON'T GUESS' and 'COMPUTE, DON'T DESCRIBE' exhortation sections and the 7-item KEY PRINCIPLES list are padding that applies to many skills rather than protein-design-specific knowledge.

4 / 5

Actionability

Concrete tool names, exact parameter names ('diffusion_steps', 'pdb_string', 'sequence'), a Common Parameter Mistakes table, and numeric thresholds (pLDDT >85, 40 RPM) give mostly executable guidance, though no copy-paste code blocks appear in the body itself.

4 / 5

Workflow Clarity

A clear 6-phase sequence plus a validation phase and a completeness checklist provide most checkpoints, but explicit validate→fix→regenerate feedback loops are not stated for failed designs even though this is a batch operation.

4 / 5

Progressive Disclosure

The Reference Files section lists one-level-deep companions (DESIGN_PROCEDURES.md, TOOLS_REFERENCE.md, EXAMPLES.md, CHECKLIST.md, design_templates.md), but no bundle files are actually present in the review directory to verify the references resolve, and the links are a flat bottom-of-file list rather than inline per-section pointers.

4 / 5

Total

16

/

20

Passed

Description

92%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A strong, specific description that pairs a concrete method list with an explicit 'Use for' trigger clause and named tools, making it both actionable and distinct. Slight room to add a few more natural synonyms (e.g. 'antibody design', 'de novo enzymes') for fuller trigger coverage.

DimensionReasoningScore

Specificity

Lists multiple concrete actions — 'RFdiffusion backbone generation, ProteinMPNN sequence design, structure validation (pLDDT, pTM, MPNN scores)' — with comprehensive coverage of the design pipeline rather than vague abstraction.

5 / 5

Completeness

Explicitly answers both what ('AI-guided de novo protein design — ...') and when ('Use for designing ... before experimental validation') with concrete trigger phrases.

5 / 5

Trigger Term Quality

The 'Use for designing therapeutic protein binders, novel scaffolds, enzyme variants, and miniprotein/protein-interface design' clause gives good natural-term coverage, but misses some common synonyms users might say.

4 / 5

Distinctiveness Conflict Risk

The named toolchain (RFdiffusion, ProteinMPNN) and 'therapeutic protein binders' niche give it a clear, distinct trigger surface with minimal overlap risk against generic skills.

5 / 5

Total

19

/

20

Passed

Validation

100%

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository
mims-harvard/ToolUniverse
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.