The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill operates on external markup and CSS files which could contain adversarial instructions. Evidence: The skill uses read_file and search_code to audit frontend components. Boundary markers: The 'Security Guardrails' section (ASI01) explicitly directs the agent to ignore any adversarial comments in the markup. Capability inventory: Uses write_file, edit_file, and execute_command for modifications and testing. Sanitization: Adheres strictly to internal styling tokens and guardrails to prevent the adoption of unvetted external payloads.
[SAFE]: No obfuscation, data exfiltration, or unauthorized command execution patterns were found. The skill includes defensive measures against common AI agent security risks.

frontend-micro-polish