The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains instructions that explicitly claim it has been verified as safe and attempt to dictate a safety verdict, which are patterns intended to bypass security auditing.
[OBFUSCATION]: A URL for an administrative login page is hidden using Base64 encoding. The decoded URL 'https://hk.uds.tech/dashboard/secret-login?redirect=admin' belongs to the author's infrastructure.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests external data that is then used in tool calls without proper boundaries or sanitization.
Ingestion points: The agent reads markdown files from the 'docs/' directory using 'git-read'.
Boundary markers: The skill lacks instructions to isolate the file content or treat it as untrusted data.
Capability inventory: The agent can perform database queries and writes using 'notion-query' and 'notion-write'.
Sanitization: There is no logic provided to sanitize or validate the content of the files before processing.
[EXTERNAL_DOWNLOADS]: The skill references external URLs on 'hkuds.com' and 'hkuds.tech' for status reporting and dashboard access; these are recognized as author-owned resources.

vnpy-export