accessibility-testing
Pass
Audited by Gen Agent Trust Hub on Mar 12, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill underwent a complete security audit against ten threat categories, including prompt injection, data exfiltration, and obfuscation, with no malicious patterns detected.
- [COMMAND_EXECUTION]: The skill includes Python scripts in the
scripts/directory that utilizesubprocess.callto orchestrate internal format conversion logic. These operations are implemented securely by using argument lists and the currentsys.executable, which avoids common shell injection vulnerabilities. - [DATA_EXFILTRATION]: All file input/output operations are restricted to the local filesystem for the purpose of saving and converting test reports. No evidence of unauthorized network activity, hardcoded credentials, or exfiltration of sensitive files was found.
- [PROMPT_INJECTION]: The system prompt provided in
prompts/accessibility-testing.mdis well-structured, establishing an expert persona for accessibility testing without attempting to override safety guardrails or bypass established agent constraints.
Audit Metadata