pi-goal
Pass
Audited by Gen Agent Trust Hub on Aug 25, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill handles user-supplied goal objectives that are subsequently interpolated into hidden agent prompts to maintain goal persistence. • Ingestion points: User input via the objective parameter in the create_goal tool and the /goal command, handled in src/index.ts. • Boundary markers: The objectives are enclosed in XML tags and are preceded by instructions to treat the data as untrusted task context rather than high-priority system instructions, as seen in src/goal/prompt.ts. • Capability inventory: The skill's capabilities are limited to goal metadata management (local file I/O) and updating the TUI status display. It lacks network access or shell execution tools. • Sanitization: Input is sanitized using XML entity escaping through the escapeXmlText function in src/goal/prompt.ts before interpolation.
- [SAFE]: No malicious patterns such as credential theft, remote code execution, or unauthorized persistence were found. The skill follows established patterns for Pi extensions and includes comprehensive tests.
Audit Metadata