baidu-drive

Pass

Audited by Gen Agent Trust Hub on Oct 30, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The install.sh script downloads the bdpan CLI installer from Baidu's CDN (issuecdn.baidupcs.com). The update.sh script fetches skill update packages from pan.baidu.com. These are recognized as official vendor resources.
  • [REMOTE_CODE_EXECUTION]: The install.sh script executes the downloaded bdpan installer. This process is protected by hardcoded SHA256 checksum verification within the script to ensure the integrity of the downloaded binary before execution.
  • [COMMAND_EXECUTION]: The instructions in SKILL.md require the agent to interpolate raw user input (--session-input) and session IDs into shell commands. To mitigate command injection risks, the skill explicitly instructs the agent to wrap user input in shell single quotes.
  • [INDIRECT_PROMPT_INJECTION]: The skill facilitates the reading of external file contents and search snippets from Baidu Drive, creating a surface for indirect prompt injection.
  • Ingestion points: The bdpan get-content and bdpan search-content commands retrieve external data that is then displayed to the agent as documented in SKILL.md.
  • Boundary markers: The skill relies on natural language instructions to warn the agent that snippets are content and not commands, rather than using structured delimiters.
  • Capability inventory: The skill utilizes the Bash tool to interact with the filesystem and network through the bdpan CLI.
  • Sanitization: No automated sanitization or filtering of the retrieved file content is performed before it is processed by the agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 30, 2026, 01:43 PM