plugin-structure
Pass
Audited by Gen Agent Trust Hub on Nov 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill documents an architecture for building plugins that process untrusted data from a user's workspace, such as source code and configuration files, creating a surface for indirect prompt injection.
- Ingestion points: Example components like the 'code-reviewer' agent and linting hooks ingest project files for analysis (e.g., 'examples/standard-plugin.md').
- Capability inventory: The documented plugin system allows for shell command execution via hooks and service integration through MCP servers.
- Boundary markers: The provided examples do not explicitly show the use of boundary markers or instructions to ignore embedded commands when processing files.
- Sanitization: The example scripts perform basic analysis but do not include explicit sanitization of the content being processed.
Audit Metadata