interactive-component-workbench
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is primarily instructional and promotes security through isolation and strict input validation. It mandates the use of dedicated scratch routes and immutable mock fixtures, preventing production contamination and ensuring a stable testing environment. The 'Security Guardrails' section provides explicit instructions to prevent common vulnerabilities like Remote Code Execution (RCE) and memory poisoning.
- [INDIRECT_PROMPT_INJECTION]: The skill architecture handles external inputs (URL search parameters) and mock data fixtures, which are identified as potential attack surfaces. However, these are secured through specific defensive requirements.
- Ingestion points: URL search parameters (
?__state,?__variant) and mock component fixtures defined in the workbench code. - Boundary markers: The skill requires mapping state and variant slugs through a fixed allowlist of permitted fixture names, ensuring only known states are rendered.
- Capability inventory: The skill environment has access to file system operations (
write_file,edit_file) and command execution (execute_command) as specified in the frontmatter configuration. - Sanitization: The instructions explicitly forbid interpolating raw parameter values into markup and emphasize treating all fixture content as untrusted data.
Audit Metadata