The Agent Skills Directory

[SAFE]: The skill implements robust internal safety protocols. It includes a dedicated references/adversarial-checks.md file that requires the agent to screen all user requests for legal conflicts, intellectual property issues (such as employer-confidential information), and regulated industry violations (like PHI handling). This defensive design significantly reduces the risk of the agent being misused for unethical or illegal activities.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves processing user-provided professional backgrounds and constraints. While this presents an injection surface, the skill mitigates this by enforcing a strict 'feasibility gate' and 'adversarial checks' workflow. These instructions command the agent to reject unrealistic or harmful requests and prioritize legal compliance over user preference, adhering to security best practices for handling untrusted input.
[SAFE]: Analysis of the skill instructions and referenced files shows no evidence of remote code execution, external package dependencies, persistence mechanisms, or unauthorized privilege escalation. The workflow focuses on analytical output and structured data generation without invoking dangerous system commands or network exfiltration patterns.

differentiated-opportunity-engine