Security audit
Security checks for vulnerabilities and agentic risk
The skill is a medical-device reporting aid with no executable payload, but it requests broad command, write, and web access while describing itself as a pure documentation tool for sensitive adverse-event data.
Review before installing. Do not provide identifiable patient, reporter, company-confidential, or regulated adverse-event details unless you have confirmed where the data will be processed and logged. The publisher should either remove Bash/Write/WebFetch for a documentation-only skill or clearly document why those tools are needed and require explicit user approval for external calls or file writes.
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
# pms-vigilance-intake · 医疗器械不良事件报告路径判定器
[](https://github.com/zhaoxinghua09-cell/lgd-theory)
The manifest advertises a zero-dependency, pure decision-support/documentation skill, yet it is granted Bash, Write, and WebFetch, which materially expand its capability to execute commands, modify files, and access remote resources. This mismatch can mislead operators and downstream policy engines into trusting a skill that has broader side-effect potential than disclosed.
The documentation explicitly claims there are no local scripts or execution needs, but Bash is permitted in the manifest. Even if unused, this creates an avoidable execution path that could be abused by prompt injection, future edits, or operator misunderstanding.
The usage sections frame the skill as pure documentation, while the risk rationale acknowledges API calls, file read/write, and local execution. In a regulated medical-reporting context, this inconsistency is dangerous because users may supply sensitive adverse-event data without realizing it could be processed through broader-capability tooling.
This skill handles adverse-event reports, which may contain sensitive personal, clinical, or commercially sensitive information, yet it lacks a clear warning about sensitive-data handling and possible backend/API processing. In a medical-device vigilance workflow, inadequate disclosure increases the risk of improper submission of regulated data to unauthorized processing paths.
Line L27 states '中文文档见 SKILL.md', indicating the skill documentation is provided in Chinese, while the README otherwise mixes English and Chinese and does not offer a user language preference or alternative locale. This can create a language/locale policy concern because the skill appears to default users into a specific language without explicit opt-in or a documented region-specific justification.