tooluniverse-admet-prediction

Comprehensive ADMET (Absorption, Distribution, Metabolism, Excretion, Toxicity) profiling for drug candidates. Integrates ADMET-AI predictions, SwissADME drug-likeness, PubChemTox experimental toxicity, ChEMBL clinical data, Lipinski rule-of-five, and CYP interaction data. Use for drug-likeness assessment, BBB penetration, bioavailability, hepatotoxicity prediction, ADME/PK profiling, or screening compound libraries before lab testing.

Quality

88%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Quality

Content

85%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 highly actionable, well-sequenced workflow with exact tool calls, argument-type distinctions, fallbacks, and validation checkpoints throughout. Its main weaknesses are modest: some textbook pharmacology explanations could be trimmed, and the ~310-line monolithic body would benefit from offloading reference material (endpoint details, interpretation tables) into references/ files.

Suggestions

Move the stable reference material — the toxicity endpoint descriptions in Phase 4, the physicochemical ideal-range table, and the verdict-flag rules — into a references/ file (e.g., references/interpretation.md), keeping SKILL.md as a lean workflow overview with clearly signaled one-level-deep links.

Trim textbook pharmacology exposition (e.g., '~75% of drugs are metabolized by CYP enzymes', hERG/QT-prolongation background, DILI withdrawal examples) down to the verdict-relevant flags, since Claude already knows the underlying biology.

Show the tool call for resolving a compound's ChEMBL ID in Phase 5 (currently 'if drug has ChEMBL ID' with no step to find it) and the exact SwissADME call backing the Phase 3 step-5 pharmacokinetics cross-validation.

Dimension	Reasoning	Score
Conciseness	The body is dense and mostly efficient — exact tool signatures, per-phase fallbacks, and interpretation thresholds ('LD50 < 50 mg/kg (FAIL: GHS 1-2)') earn their tokens. However, minor instances of over-explanation of textbook pharmacology ('~75% of drugs are metabolized by CYP enzymes', 'hERG... Causes QT prolongation and fatal cardiac arrhythmia') could be trimmed, matching the 'efficient with minor over-explanation' anchor rather than the lean level-5 anchor.	4 / 5
Actionability	Fully executable guidance: copy-paste-ready calls with exact tool names and argument types ('ADMETAI_predict_BBB_penetrance(smiles=["<SMILES>"])'), explicit list-vs-string argument distinctions, install command ('uv pip install \'tooluniverse[ml]\''), fallbacks per phase, and a specified output artifact (13-category pass/warn/fail scorecard). Common cases — name input, SMILES input, missing ml extra — are all covered.	5 / 5
Workflow Clarity	A 5-phase pipeline with an explicit diagram, numbered steps per phase, explicit validation and error-recovery feedback loops (fallback paths when ADMETAI import fails; 'Only treat output as a failure if the tool returns an error field or no predictions'; expected-console-noise guidance preventing false retries), and a mandatory completeness checklist before reporting. This matches the level-5 anchor including feedback loops and checklists.	5 / 5
Progressive Disclosure	A single ~310-line file with good section headers but no bundle files at all — everything is inline, including material that would fit separate references (the toxicity endpoint encyclopedia in Phase 4, the physicochemical ideal-range table, the verdict-flag rules). Structure is good, so it is above the level-2 anchor, but content that should be split out is inline, matching the level-3 anchor.	3 / 5
	Total	17 / 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 description: specific, comprehensive, third-person, with an explicit 'Use for' trigger clause. The only weakness is a handful of missing natural trigger synonyms (toxicity prediction, LD50, SMILES) that users commonly use when asking about compound safety.

Suggestions

Add commonly-used trigger phrases such as 'toxicity prediction', 'LD50/hERG liability', and 'SMILES' to the 'Use for' clause so users phrasing questions around those terms match the skill.

Mention novel-compound screening by SMILES explicitly (e.g., 'SMILES-based toxicity prediction') since the skill accepts SMILES input but the description only implies it.

Dimension	Reasoning	Score
Specificity	The description lists multiple specific, named actions — 'Integrates ADMET-AI predictions, SwissADME drug-likeness, PubChemTox experimental toxicity, ChEMBL clinical data, Lipinski rule-of-five, and CYP interaction data' — covering prediction, rule-based, experimental, and clinical evidence tiers. Coverage is comprehensive with no minor gaps, matching the level-5 anchor rather than level 4.	5 / 5
Completeness	It explicitly answers both what ('Comprehensive ADMET... profiling for drug candidates. Integrates...') and when ('Use for drug-likeness assessment, BBB penetration, bioavailability, hepatotoxicity prediction, ADME/PK profiling, or screening compound libraries before lab testing'), with concrete trigger phrases in third-person voice.	5 / 5
Trigger Term Quality	'drug-likeness assessment, BBB penetration, bioavailability, hepatotoxicity prediction, ADME/PK profiling, screening compound libraries' are natural user phrases with good coverage. A few natural terms users would say are missing — 'toxicity prediction', 'LD50', 'SMILES' — so it falls between the good-coverage (4) and comprehensive-synonym (5) anchors, closer to 4.	4 / 5
Distinctiveness Conflict Risk	The ADMET/drug-candidate-profiling niche is clear, with named databases (ADMET-AI, SwissADME, PubChemTox, ChEMBL) and domain-specific triggers that would not fire for general chemistry, documents, or data skills. Minimal conflict risk.	5 / 5
	Total	19 / 20 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-admet-prediction/SKILL.md
Commit: 164aaf1

Reviewed: 6 days 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.