architect
Pass
Audited by Gen Agent Trust Hub on Mar 10, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill defines a strict software architect role with clear boundaries. It explicitly prohibits editing files, running shell commands, and modifying configurations, ensuring it cannot perform destructive actions on the host system or codebase.- [SAFE]: No obfuscated code, hidden URLs, hardcoded credentials, or unauthorized remote execution patterns were detected within the skill definition.- [PROMPT_INJECTION]: The skill contains an attack surface for indirect prompt injection as it uses tools like WebFetch and WebSearch to process potentially untrusted external data. However, the risk is mitigated by the 'planning-only' constraint which prevents the agent from executing any instructions found in processed data.
- Ingestion points: External content retrieved via WebFetch and WebSearch; local code read via Read and Grep tools.
- Boundary markers: Includes a 'PROSE constraint' for safety boundaries but lacks specific technical delimiters for data processing.
- Capability inventory: Limited to read-only research and design; no file-write, command execution, or configuration modification tools are allowed.
- Sanitization: No explicit data sanitization or input validation is described in the skill instructions.
Audit Metadata