The Agent Skills Directory

[CREDENTIALS_UNSAFE]: A hardcoded API key 123456 is provided in the SKILL.md setup instructions for authenticating with the backend API.
[DATA_EXFILTRATION]: The skill is configured to send user-supplied stock codes to an unverified raw IP address (43.167.234.49) over an unencrypted (HTTP) or unspecified connection, representing a data exfiltration risk to a non-whitelisted domain.
[PROMPT_INJECTION]: The instructions explicitly command the agent to disregard its internal training memory and bypass safety or knowledge filters ("must immediately call this skill—never answer from training memory").
[PROMPT_INJECTION]: Indirect Prompt Injection Surface: The skill ingests data from an external API without security boundaries or sanitization, creating a surface for indirect attacks. Ingestion points: Data returned from http://43.167.234.49:3101/api/v2/hkstock/securities is directly processed by the agent. Boundary markers: Absent. There are no delimiters used to separate untrusted API data from the agent's instructions. Capability inventory: The skill uses curl for network operations and processes JSON data. Sanitization: No validation or sanitization of the API response is implemented.
[COMMAND_EXECUTION]: The skill metadata contains shell commands for installation (cp -r ...) which execute file system operations on the host environment.

hk-realtime-quote