tooluniverse-adverse-event-detection

Detect and analyze adverse drug event signals using FDA FAERS reports, drug labels, and disproportionality statistics (PRR, ROR, IC). Generates quantitative safety signal scores (0-100) with evidence grading. Use for post-market surveillance, pharmacovigilance, drug safety assessment, regulatory submissions, and detecting rare AE signals not visible in clinical trials.

Quality

71%

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

Fix and improve this skill with Tessl

tessl review fix ./plugin/skills/tooluniverse-adverse-event-detection/SKILL.md

Quality

Content

50%Weight 40%Scale 1-3

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

A well-organized, domain-rich skill with a clear phased workflow and specific tool/criteria guidance, but it is held back by missing bundle files, no inline executable code, and absent validation checkpoints for batch operations. The progressive-disclosure structure is sound in concept but broken in practice because every referenced detail file is absent.

Suggestions

Create the four referenced bundle files (`PHASE_DETAILS.md`, `REPORT_TEMPLATE.md`, `TOOL_REFERENCE.md`, `QUICK_START.md`) — currently all referenced paths 404, breaking progressive disclosure and the deferred code examples.

Add inline executable Python for the core computation (PRR/ROR/IC with 95% CI and the 0-100 Safety Signal Score) so the body is copy-paste ready rather than deferring all code to a missing file.

Insert explicit validation checkpoints between phases for the batch operations, e.g., after Phase 2 verify each AE has N>=3 and a computable CI before stratifying, and after Phase 8 confirm every signal has a T1-T4 grade before report synthesis.

Tighten redundancy: state the PRR/signal criteria once (Phase 2) and have "REASONING STRATEGY"/"Causality Assessment" reference it instead of re-stating thresholds.

Dimension	Reasoning	Score
Conciseness	The body assumes Claude's competence (no basic-concept padding) and is dense with domain-specific guidance, but it is verbose with redundancy — the PRR threshold and signal criteria appear in both "REASONING STRATEGY" and Phase 2, and "KEY PRINCIPLES" overlaps the phase summaries. Not a 3 because several sections could be tightened without losing clarity; not a 1 because it never explains concepts Claude already knows.	2 / 3
Actionability	Concrete guidance is strong — specific tool names (e.g., `FAERS_calculate_disproportionality`), exact criteria ("PRR >= 2.0 AND lower CI > 1.0 AND N >= 3"), and explicit scoring weights (0-35/0-30/0-25/0-10). Not a 3 because there is no executable code in the body and the promised code examples live only in `PHASE_DETAILS.md`, which does not exist, so key implementation details are missing.	2 / 3
Workflow Clarity	The 10-phase pipeline is clearly sequenced with a diagram, per-phase tools, and edge-case handling. Capped at 2 per the batch-operation rule: there are no explicit validate→fix→retry feedback checkpoints between phases (e.g., confirm FAERS returned usable data and valid CIs before stratifying), despite this being a multi-step batch analysis skill.	2 / 3
Progressive Disclosure	SKILL.md is well-structured with clearly signaled one-level-deep references ("See `PHASE_DETAILS.md` for full signal classification table", "See `REPORT_TEMPLATE.md` for full template"), which would normally score 3. Scored against the actual bundle: none of the four referenced files (`PHASE_DETAILS.md`, `REPORT_TEMPLATE.md`, `TOOL_REFERENCE.md`, `QUICK_START.md`) exist, so following the references dead-ends. Not a 1 because the overview itself is organized and references are one level deep, not nested.	2 / 3
	Total	8 / 12 Passed

Description

92%Weight 40%Scale 1-3

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, third-person description that clearly states concrete capabilities and provides an explicit use-when clause with natural trigger terms. The only weakness is moderate overlap risk with a general pharmacovigilance skill, which the body itself acknowledges.

Suggestions

Sharpen the trigger clause to emphasize signal detection/quantification over generic 'pharmacovigilance' to reduce overlap with a sibling PV skill (e.g., 'Use when quantifying FAERS disproportionality signals or computing a Safety Signal Score, not for general pharmacovigilance lookups').

Front-load a differentiator phrase such as 'quantitative signal detection' early so the distinction from a broader pharmacovigilance skill is visible before the generic 'drug safety assessment' trigger.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete actions — "Detect and analyze adverse drug event signals using FDA FAERS reports, drug labels, and disproportionality statistics (PRR, ROR, IC)" and "Generates quantitative safety signal scores (0-100) with evidence grading" — matching the multi-action anchor. Not a 2 because it goes beyond naming a domain to enumerate specific quantitative actions.	3 / 3
Completeness	Explicitly answers both what (detect/analyze signals, generate 0-100 scores with grading) and when (an explicit "Use for..." trigger list). Not a 2 because the when-clause is present and explicit, not merely implied.	3 / 3
Trigger Term Quality	The "Use for post-market surveillance, pharmacovigilance, drug safety assessment, regulatory submissions, and detecting rare AE signals" clause covers natural terms the target audience would say. Not a 2 because it gives broad coverage of real phrasing rather than only a single keyword, though some jargon (PRR/ROR/IC) is mixed in.	3 / 3
Distinctiveness Conflict Risk	The niche (FAERS disproportionality signal detection with PRR/ROR/IC and a 0-100 Safety Signal Score) is specific, but the broad trigger "pharmacovigilance"/"drug safety assessment" overlaps with a sibling pharmacovigilance skill the body itself differentiates from. Not a 3 because that acknowledged overlap means it could trigger for the wrong skill; not a 1 because the statistical methods and scored output carve a clear sub-niche.	2 / 3
	Total	11 / 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-adverse-event-detection/SKILL.md
Commit: 91e8040

Reviewed: 11 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.