android-material3-compose
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill requires the agent to run a bundled Python script
scripts/inspect-project.pyfor repository analysis and configuration auditing.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data by analyzing source files from the user's repository, which creates a potential surface for indirect prompt injection.\n - Ingestion points: The
inspect-project.pyscript recursively scans the project root and reads files (such as.kt,.gradle, and.toml) to detect Material 3 setup and migration patterns.\n - Boundary markers: The script output lacks explicit delimiters or instructions to the agent for handling the analysis results as untrusted content.\n
- Capability inventory: The skill uses local shell command execution and file system access capabilities.\n
- Sanitization: The tool uses regular expressions to extract metadata and returns structured summaries and file paths rather than raw file content, reducing the risk of file contents being interpreted as direct instructions by the AI.
Audit Metadata