lint
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The shell scripts used to find and check files (e.g.,
for script in $changed_scripts) are susceptible to word splitting if filenames contain spaces or special characters. While the use of[[ -f "$script" ]]prevents execution of non-existent files resulting from such splits, it represents a fragile pattern for handling untrusted directory content. - [INDIRECT_PROMPT_INJECTION]: The skill analyzes shell scripts and GitHub Action workflows from the repository, which can contain instructions intended to influence the agent when it attempts to "Fix the issues."
- Ingestion points: File content from shell scripts (
*.sh) and workflow files (.github/workflows/*.yml) identified viagit diff. - Boundary markers: None. The content of the files is processed directly by linting tools without delimiters or instructions to ignore embedded commands.
- Capability inventory: The skill enables the agent to run command-line tools (
git,shellcheck,actionlint) and provides instructions for the agent to modify the source code of the project. - Sanitization: No validation or filtering is performed on the content of the analyzed files prior to processing or acting upon the results.
Audit Metadata