The Agent Skills Directory

[SAFE]: The skill performs expected business operations for CRM and email platform management without any detectable malicious intent or harmful command execution.
[DATA_EXFILTRATION]: The skill reads and writes to a specific directory (~/.claude-marketing/) dedicated to the application's brand profiles and execution logs. This is standard behavior for maintaining state and configuration in a marketing automation context.
[PROMPT_INJECTION]: The skill processes data from external sources (CRM contacts, email engagement) and local JSON configuration files, representing a theoretical surface for indirect prompt injection.
Ingestion points: profile.json, _manifest.json, and connected CRM/Ad platform data.
Boundary markers: Not explicitly defined in the provided process logic.
Capability inventory: Reading configuration files, writing to segment-log.json, and interacting with CRM/Marketing APIs via secondary agents (crm-manager, email-specialist, media-buyer).
Sanitization: No explicit sanitization or validation steps for external data are described in the instructions, which is common for high-level orchestration skills.

segment-audience