architect-role-skill
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONREMOTE_CODE_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute standard shell commands (find, cat, git) to perform codebase discovery. These commands are limited to reading project metadata and file structures within the local environment.
- [INDIRECT_PROMPT_INJECTION]: The skill establishes an attack surface by ingesting content from local project files (e.g., dependency manifests) and git logs into the agent's context. This is required for its architectural analysis function. Ingestion points: SKILL.md (Step 2.1 Discovery Phase) reads external file contents and shell output. Boundary markers: None identified. Capability inventory: Uses find, cat, and git for read-only access to local repository data. Sanitization: None identified.
- [REMOTE_CODE_EXECUTION]: The skill's resource section references local shell scripts (scripts/analyze_codebase.sh, scripts/dependency_audit.sh). No patterns of remote code execution, such as downloading and piping scripts from external URLs, were found.
Audit Metadata