The Agent Skills Directory

[SAFE]: The skill is entirely composed of markdown instructions and evaluation data. It provides a framework for analyzing user onboarding flows without implementing any functional code.- [PROMPT_INJECTION]: Analysis of the instructions shows no evidence of jailbreak attempts, safety bypasses, or instructions to ignore previous system prompts. The 'expert' persona is strictly bounded to the marketing domain.- [DATA_EXFILTRATION]: The skill instructions include checks for a local file named '.agents/product-marketing-context.md'. This is a legitimate pattern for gathering project-specific context and does not involve any external network calls or unauthorized data transmission.- [REMOTE_CODE_EXECUTION]: No commands related to script execution, package installation (pip, npm), or dynamic code evaluation were found in any of the analyzed files.

onboarding-cro