tuzi-copy-polish
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes shell commands (
test -f) in Step 0.1 to detect the presence of configuration files (EXTEND.md) in the current project directory and the user's home directory. - [COMMAND_EXECUTION]: In Step 3, the skill performs a file write operation by appending the optimized copy to the end of a file path provided by the user. This operation lacks path validation, which could be misused to append text to sensitive system or configuration files if an attacker provides a malicious path.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from files or inline text to perform the polishing task, creating a surface for indirect prompt injection.
- Ingestion points: File content reading in Step 1 and user-provided inline text.
- Boundary markers: The skill does not define explicit delimiters to separate the input data from the agent's instructions during the rewriting process.
- Capability inventory: Shell command execution (file existence checks) and file system write (append) capabilities.
- Sanitization: There is no evidence of input sanitization or validation of the target file paths.
Audit Metadata