gstack-openclaw-office-hours
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from the local workspace and git history, which are external ingestion points. While this creates a surface for indirect instructions, the skill implements a strict policy against code execution.
- Ingestion points: Workspace files and git logs (Phase 1).
- Capability inventory: Restricted to design documentation and writing to the
memory/directory; explicitly barred from code execution by a "HARD GATE" instruction. - Boundaries and Sanitization: No specific delimiters are used, but the behavioral constraint provides a safety barrier against execution of malicious payloads found in the workspace.
- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill reads project documentation and git history to provide relevant context. This local access is required for its function and no network exfiltration patterns were detected.
- [NO_CODE]: The instructions contain a "HARD GATE" that strictly prevents the agent from performing any implementation, code generation, or project scaffolding tasks, ensuring its output remains limited to design documentation.
Audit Metadata