cloning-project
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill relies on shell commands to gather, move, and archive files, including 'cat', 'ls', 'cp', 'mkdir', and 'zip'. These operations are performed on user-controlled data paths like '/mnt/user-data/uploads/'.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted project instructions and knowledge files which may be influenced by third-party content.
- Ingestion points: The skill extracts project instructions directly from the agent's context window and reads files from the '/mnt/user-data/uploads/' directory.
- Boundary markers: The skill uses a single-quoted heredoc ('INSTRUCTIONS') when writing the extracted instructions to a file, which is a security best practice that prevents the shell from interpreting variables or commands within the text.
- Capability inventory: Includes the ability to write to the file system, create directories, and generate zip archives via 'bash'.
- Sanitization: No explicit validation or filtering is applied to the content of the project instructions or the knowledge files before they are processed and bundled for export.
Audit Metadata