The Agent Skills Directory

[SAFE]: The skill includes a specific security instruction to treat external inputs—such as journalist lists, coverage exports, and inbound email replies—as untrusted, warning the agent never to follow instructions embedded within them.
[SAFE]: The skill implements a 'least privilege' data model where it only writes to 'candidate' files (e.g., memory/launch-registry/candidates.md) rather than modifying authoritative records directly, preventing accidental or malicious corruption of core project state.
[SAFE]: No obfuscation, prompt injection attempts, or unauthorized network operations were detected. External references point to local project scripts and standard marketing frameworks.

press-media-relations