architecture-md-builder
Pass
Audited by Gen Agent Trust Hub on May 5, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to use standard CLI tools like
wc -lfor line count verification andcpfor template management. It also references the execution of a Python script from a related local skill (exa-search) to perform repository research. These operations are restricted to the local environment and are consistent with the skill's purpose of codebase analysis. - [EXTERNAL_DOWNLOADS]: The skill suggests using an external search tool to find exemplary documentation on GitHub. This is an informational search and does not involve the automated download or execution of untrusted remote code.
- [DATA_EXFILTRATION]: No sensitive data access or exfiltration patterns were identified. The exploration phase is focused on structural metadata (abstractions, file sizes, directory trees) and does not target sensitive configuration files like credentials or environment secrets.
- [PROMPT_INJECTION]: The instructions provided to sub-agents are task-specific and do not contain bypass markers, role-play attempts, or commands to ignore safety guidelines.
Audit Metadata