plugin-best-practices
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_CONTEXT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes local Python scripts bundled with the plugin, such as
validate-plugin.pyandaudit-bare-paths.py, to verify plugin structure and frontmatter compliance.\n- [INDIRECT_PROMPT_INJECTION]: The skill analyzes third-party plugin components (manifests and scripts), representing an indirect prompt injection surface. This ingestion is the primary function of the validation workflow.\n - Ingestion points: External plugin paths passed to scripts as instructed in
SKILL.md.\n - Boundary markers: None defined for the validation logic itself.\n
- Capability inventory: Filesystem access and execution of validation scripts via Python.\n
- Sanitization: Not explicitly discussed in the provided reference materials.\n- [DYNAMIC_CONTEXT_INJECTION]: The documentation explains the use of the
!command`` syntax for dynamic context injection, allowing developers to include shell command output in plugin prompts.
Audit Metadata