help
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a local Python script (
scripts/plugin-metadata.py) at runtime to fetch current version, asset counts, and connector status. This execution uses the platform-provided${CLAUDE_PLUGIN_ROOT}environment variable to locate the script. - [INDIRECT_PROMPT_INJECTION]: The skill implements an intent-routing feature that takes user-supplied goals to recommend sequences of other skills.
- Ingestion points: User-provided strings passed via the
--intent "<goal>"argument or natural language queries inSKILL.md. - Boundary markers: None are explicitly defined to isolate the user-provided goal string from the internal matching logic.
- Capability inventory: The skill can execute the plugin's metadata script, read the
skills-index.jsonfile, and list various slash commands available within the plugin inSKILL.md. - Sanitization: No specific sanitization or validation of the goal input is described before it is matched against the skill index descriptions.
Audit Metadata