aitodo
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes instructions directly from source code comments (AITODO).
- Ingestion points: The skill ingests untrusted data from the local repository using
git grep AITODOinSKILL.md. - Boundary markers: There are no delimiters or warnings instructed to the agent to treat the content of these comments as potentially malicious or untrusted data.
- Capability inventory: The skill allows the agent to perform arbitrary code modifications and deletions in the local file system.
- Sanitization: No sanitization, validation, or filtering of the comment content is performed before the agent implements the requested changes.
- [COMMAND_EXECUTION]: The skill relies on shell commands (
git grep) to locate task definitions within the codebase. While standard for developer tools, this command identifies the data source for the indirect prompt injection surface.
Audit Metadata