tooluniverse-protein-sae-variant-interpretation

Interpret a missense variant via ESMC-6B Sparse Autoencoder (SAE) feature activations. For a given protein + variant, computes which interpretable SAE features (catalytic, ligand-binding, PTM, structural motif, domain, etc.) are lost or gained at the mutation site. Use when standard pathogenicity scores (AlphaMissense, ClinVar) say a variant is damaging but you need a MECHANISTIC explanation — e.g. 'why is this variant LoF?' Complements (does not replace) variant-interpretation and variant-to-mechanism skills, which focus on ACMG classification or regulatory mechanism.

Quality

82%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Low

Low-risk findings worth noting

Quality

Content

65%

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

The body is highly actionable with executable code and a clearly sequenced, validated core workflow, but it runs long with overlapping workflow variants and keeps everything inline rather than using progressive disclosure. Adding validation guidance for batch operations and offloading the reference tables to bundled files would lift the weaker dimensions.

Suggestions

Add an explicit validation/recovery step for the batch path (ESM_score_variant_sae_batch) — e.g. how to handle a ref_aa mismatch or partial failures across the 19 alternates — so batch operations match the main workflow's checkpoint rigor.

Consolidate the overlapping workflow variants (fullest one-call, raw-delta-only, batch, long path) into a single decision tree or trim redundant commentary to tighten token usage.

Move the interpretation table, cross-validation pattern, and reporting format into a bundled reference file (e.g. references/INTERPRETATION.md) referenced one level deep, leaving SKILL.md as a lean overview.

Dimension	Reasoning	Score
Conciseness	Mostly efficient and assumes Claude's competence — it adds only domain-specific, non-obvious content (Forge tool params, return shapes, credit costs, caveats) rather than explaining basics. Not a 3 because the ~240-line body presents overlapping workflows (fullest one-call, raw-delta, batch, long path) and extensive tables that could be tightened.	2 / 3
Actionability	Fully executable, copy-paste-ready code throughout (UniProt_search, ESM_explain_variant_mechanism, ESM_score_variant_sae_disruption, the feature_to_window_sum function, and a batch example) with concrete params and documented return shapes.	3 / 3
Workflow Clarity	The main 5-step workflow is clearly sequenced with an explicit residue-validation checkpoint and error feedback, but the batch-operation path (ESM_score_variant_sae_batch) lacks explicit validation/recovery steps, which caps the score per the destructive/batch-operations guideline.	2 / 3
Progressive Disclosure	A single ~240-line file with no bundle files (references/scripts/assets absent), so content that could be split out (interpretation table, cross-validation pattern, reporting template) is all inline. Sections are well-organized, but it is neither a small under-50-line skill nor does it use one-level-deep external references.	2 / 3
	Total	9 / 12 Passed

Description

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.

The description is specific, trigger-rich, and complete with an explicit 'Use when' clause and clear boundary-setting against sibling skills. It avoids vagueness and over-claims while staying concise.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete actions ('Interpret a missense variant', 'computes which interpretable SAE features ... are lost or gained') and enumerates specific feature categories (catalytic, ligand-binding, PTM, structural motif, domain).	3 / 3
Completeness	Explicitly answers both what it does (computes lost/gained SAE features) and when to use it via an explicit 'Use when standard pathogenicity scores ... say a variant is damaging but you need a MECHANISTIC explanation' clause.	3 / 3
Trigger Term Quality	Strong coverage of natural domain terms users would say: 'missense variant', 'pathogenicity scores', 'AlphaMissense', 'ClinVar', 'damaging', 'LoF', and 'MECHANISTIC explanation'.	3 / 3
Distinctiveness Conflict Risk	Carves out a clear niche (SAE-based mechanistic explanation) and explicitly disambiguates from neighboring skills: 'Complements (does not replace) variant-interpretation and variant-to-mechanism skills, which focus on ACMG classification or regulatory mechanism.'	3 / 3
	Total	12 / 12 Passed

Validation

100%

Warnings & errors only

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

Validation — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository: mims-harvard/ToolUniverse
Path: plugin/skills/tooluniverse-protein-sae-variant-interpretation/SKILL.md
Commit: 089eb8e

Reviewed: about 7 hours ago

Table of Contents

Discovery Implementation Validation

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.