express-bus-booking
Fail
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: HIGHREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill instructs the agent and user to use
npx -y @nomadamas/k-skill@0to fetch primary instructions, list files, and update the skill. This pattern allows the maintainer of the@nomadamas/k-skillpackage to execute arbitrary code or change the agent's behavior at runtime without local verification. - Evidence:
npx -y @nomadamas/k-skill@0 instruct express-bus-bookinginSKILL.md. - Evidence:
npx -y @nomadamas/k-skill@0 exec express-bus-booking scripts/kobus_express_booking.py ...ininstruction.md. - [EXTERNAL_DOWNLOADS]: The skill relies on an external package from a non-standard source (
@nomadamas/k-skillon npm) that is not affiliated with established trusted organizations. - [COMMAND_EXECUTION]: The instructions require the execution of shell commands (
npx) to perform core functionalities, including reading documentation and executing the helper Python script. - [DYNAMIC_EXECUTION]: The Python script
scripts/kobus_express_booking.pyexplicitly disables SSL certificate verification and lowers the security level for ciphers, which is a dangerous practice for a tool handling booking and payment-related flows. - Evidence:
ctx = ssl._create_unverified_context()andctx.set_ciphers("DEFAULT@SECLEVEL=1")inscripts/kobus_express_booking.py. - [INDIRECT_PROMPT_INJECTION]: The skill possesses a significant attack surface by ingesting untrusted data from an external website (
www.kobus.co.kr) and interpolating it into a generated HTML helper file. - Ingestion points: Bus terminal codes, timetable data, and seat identifiers fetched from KOBUS HTTP responses in
scripts/kobus_express_booking.py. - Boundary markers: None. There are no instructions to the agent to disregard instructions embedded in the fetched bus route data.
- Capability inventory: The skill has network access (
urllib.request), file writing capabilities (Path.write_text), and command execution authority (npx). - Sanitization: While the script uses
html.escapefor HTML generation, there is no validation or sanitization of the data retrieved from the external site before it is used to construct a local auto-submitting form.
Recommendations
- AI detected serious security threats
Audit Metadata