Security audit
Security checks for vulnerabilities and agentic risk
This is a coherent Hong Kong MTR ETA skill that uses disclosed transit data sources and local station-cache files for its stated purpose.
Before installing, confirm you are comfortable with the skill making outbound requests to the disclosed Hong Kong MTR and DATA GOV HK endpoints and refreshing a local station CSV inside the skill directory. No evidence of secret handling, unrelated file access, or hidden execution was found.
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
"Line Code","Direction","Station Code","Station ID","Chinese Name","English Name","Sequence"
"AEL","DT","AWE","56","博覽館","AsiaWorld-Expo",1.00
"AEL","DT","AIR","47","機場","Airport",2.00
"AEL","DT","TSY","46","青衣","Tsing Yi",3.00
The skill invokes Python scripts and documents network/file access, but it does not declare enforceable tool scope such as permissions or allowed-tools. That creates a policy gap: a host agent may permit broader file reads or outbound network use than intended, so the skill's safety relies only on prose instructions rather than technical restriction.
The README states that output language is tc or en, which imposes a narrow locale/language constraint in the skill's natural-language interface description. Because the file does not explain that this is a user-selectable option driven by data-source limitations or region-specific scope, it risks violating the policy against forced language constraints without clear opt-in or justification.
The language handling policy states that Chinese queries map to TC and English/non-Chinese queries map to EN, which imposes a locale choice automatically. This is a natural-language policy concern because the user is not explicitly offered a language preference or opt-in override in the documented behavior.
The CLI entry point sets lang to "EN" whenever the user does not provide a language argument. This forces a specific language choice by default rather than offering an explicit user selection, which matches the language/locale policy concern for natural-language behavior.
This CSV includes bilingual station names with a fixed Chinese/English locale schema across the dataset. Because SQP-3 applies to all file types and flags language/locale policy violations, the file presents a predefined locale choice without any accompanying natural-language justification or opt-in mechanism.
No suspicious patterns detected.