yao-web
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to retrieve and process data from external web sources via the
web_searchandweb_fetchtools. This content is subsequently integrated into the agent's active context. - Ingestion points: External content retrieved from URLs and search results in
SKILL.md. - Boundary markers: Absent. The instructions do not define delimiters or provide warnings to the agent to disregard potential instructions embedded within the fetched web content.
- Capability inventory: The skill utilizes the
tai toolCLI for its operations, which involves shell-level execution. - Sanitization: None specified. The skill assumes the retrieved content is safe for direct processing.
- [COMMAND_EXECUTION]: The skill uses a command-line interface (
tai) to execute its core functions. It passes JSON-formatted arguments (including user-derived search queries and URLs) directly to this tool via bash. While this is the intended design for tool interaction, it relies on the underlying platform to handle argument escaping safely.
Audit Metadata