code-audit
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the analyzed codebase (Phase 1-3 in SKILL.md), creating a surface where malicious instructions in source files could influence auditing agents. While boundary markers are present in the prompt template to delimit context, no explicit sanitization is performed on the ingested code. The skill possesses capabilities for subagent invocation, file system access, and command execution (git and npm) as defined in SKILL.md.\n- [COMMAND_EXECUTION]: Employs standard developer tools like 'git' and 'npm' for project analysis and verification. These operations are restricted to the project environment and are consistent with the skill's stated purpose.
Audit Metadata