The Agent Skills Directory

[SAFE]: The skill maintains strong operational control through 'Action-Level Phase Gates', ensuring the agent only proceeds when minimum input requirements are met and explicit user authorization is obtained for sensitive operations like file modification or external transmission.
[SAFE]: 'Forbidden Modes' are explicitly defined to prevent the agent from fabricating data, exaggerating qualifications, or misleading the user with hallucinated matching scores.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface as it processes untrusted user-provided content such as resumes and job descriptions. Ingestion points include user-provided materials like resumes and JDs. Boundary markers include the structured Routing Contract to isolate data. Capability inventory is limited to task coordination without direct file-write or network access. Sanitization includes constraints that prevent the model from interpreting embedded instructions as commands.

career-hub