Security audit

lab-unit-harmonization

Security checks for vulnerabilities and agentic risk

Overview

This skill is not malware, but it should be reviewed because its lab-cleaning instructions can silently drop or alter clinical measurements.

Review this skill carefully before using it on real patient or research data. It should require explicit source units or trusted source mappings, preserve original values, log every conversion or dropped record, keep full-precision numeric canonical data, and treat range checks as validation warnings rather than automatic conversion or clamping rules.

A.I.G

Vulnerability Patterns

Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally

Findings (3)

T09 · Insecure Skill Coding Practices

Warning

Location: SKILL.md:119
Finding: Ambiguous comma parsing can silently corrupt clinical measurements
Content: View full analysis
Remediation: View remediation

T09 · Insecure Skill Coding Practices

Error

Location: SKILL.md:140
Finding: Range-based unit inference can silently apply an incorrect conversion
Content: View full analysis
Remediation: View remediation

T09 · Insecure Skill Coding Practices

Warning

Location: SKILL.md:202
Finding: Uniform two-decimal formatting causes clinically significant precision loss
Content: View full analysis
Remediation: View remediation
`, and below-detection-limit indicators. - Store both the original source value and any display-formatted value. - Keep analytical columns numeric rather than replacing them with strings. - Add boundary tests for troponins, urine specific gravity, pH, and other precision-sensitive measurements. ]]>

SkillSpector

Vulnerability Patterns

MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (2)

Intent-Code Divergence

Medium

Category: Not specified by scanner
Confidence: 95% confidence
Finding: The skill instructs users to drop any row containing missing numeric values before harmonization, which is a destructive data-processing step not required for unit normalization. In healthcare analytics, this can silently remove large portions of patient data, bias cohorts, distort downstream models or clinical summaries, and potentially exclude vulnerable patients with partial lab panels.
Content: No source excerpt is available for this finding.

Intent-Code Divergence

Medium

Category: Not specified by scanner
Confidence: 98% confidence
Finding: The documentation first states that out-of-range values should be returned unchanged and explicitly says not to clamp, but later permits tolerance-based clamping near boundaries. This inconsistency can lead implementers to silently alter clinical measurements at decision thresholds, masking true abnormalities and producing non-reproducible transformations across systems.
Content: No source excerpt is available for this finding.