prevent-drift
Pass
Audited by Gen Agent Trust Hub on Nov 6, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute local
gitcommands, includinggit diffandgit ls-files, to retrieve information about the repository's state. - [DYNAMIC_EXECUTION]: The skill provides an embedded Python script to be executed via a shell heredoc (`python3
- <<'PY'`). This script performs text analysis on files tracked in the repository.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from local repository files to identify similarities, presenting an attack surface where malicious instructions in files could influence agent behavior.
- Ingestion points: Local file content is read into memory using
pathlib.Path(f).read_text()within theSKILL.mdfile. - Boundary markers: There are no explicit markers or warnings used to separate the ingested file content from the agent's instructions.
- Capability inventory: The skill can execute local shell commands (
git) and read the filesystem, but it lacks network access or file-writing capabilities. - Sanitization: The content of the files is used strictly for similarity comparison and is printed to the output; it is not used to generate executable code or subsequent system instructions.
Audit Metadata