agency-evidence-collector
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [SAFE]: The skill defines a skeptical QA agent persona focused on visual verification and evidence-based reporting. It provides a structured methodology for testing web applications.
- [COMMAND_EXECUTION]: The skill includes shell commands for local testing (
./qa-playwright-capture.sh,ls,grep,cat). These are standard diagnostic and testing commands executed within the local project environment. - [INDIRECT_PROMPT_INJECTION]: The skill processes project files (HTML, CSS, screenshots) to identify issues. While it interacts with external data, it does so for legitimate QA purposes and does not display unsafe interpolation or high-privilege capabilities that would lead to injection vulnerabilities.
- [EXTERNAL_DOWNLOADS]: The skill references an external instruction file (
ai/agents/qa.md), which appears to be a local project reference for extended documentation. No remote network downloads from untrusted sources were found.
Audit Metadata