The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from the local filesystem (memory/customer-profiles/ and memory/messaging/) to inform outreach strategies. While this involves reading external content, the risk is mitigated by a mandatory human review process where the user must approve or adapt all drafted messages before they are sent. Additionally, the skill's configuration disables autonomous model invocation.
Ingestion points: Reads from memory/customer-profiles/ and memory/messaging/ (SKILL.md).
Boundary markers: None explicitly defined for file contents.
Capability inventory: Performs file read and write operations within the memory/ directory (SKILL.md).
Sanitization: Not explicitly defined, though the workflow relies on human oversight.
[DATA_EXPOSURE]: The skill stores message iterations and outcomes in local markdown files within the memory/ folder. This is a standard state-management practice for agent skills and does not involve access to sensitive system directories or credentials.

outreach