The Agent Skills Directory

[SAFE]: The skill is entirely instructional and serves as a local audit tool for legal content. It does not contain executable code, scripts, or external dependencies that could compromise system security.- [SAFE]: The execution environment is highly restricted. The skill uses the local data residency setting, prohibits PII egress, and limits the agent to the Read tool only, which prevents unauthorized data exfiltration or command execution.- [SAFE]: Although the skill processes untrusted user input (prompts for audit), the lack of network access and restricted toolset effectively mitigates risks associated with indirect prompt injection.- [SAFE]: The skill mandates a 'Human Gate,' requiring an authorized user to review and manually execute any actions (like pasting a redacted prompt), ensuring that the agent does not operate autonomously in sensitive workflows.

privilege-preflight-en