The Agent Skills Directory

[SAFE]: All files (SKILL.md, evals/evals.json, references/ideas-by-category.md) consist of static text and structured data for marketing guidance. There is no presence of executable scripts, system commands, or external dependencies in the bundle.
[DATA_EXPOSURE_AND_EXFILTRATION]: The instructions direct the agent to read .agents/product-marketing-context.md for project context. This is a benign local file read operation used to personalize marketing advice and does not involve access to sensitive system paths or credentials.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from local configuration files. Evidence chain:
Ingestion points: .agents/product-marketing-context.md (SKILL.md)
Boundary markers: Absent
Capability inventory: None (no shell, network, or file-write capabilities detected in the skill instructions)
Sanitization: Absent
Conclusion: Given the lack of actionable capabilities, this ingestion surface poses no significant security risk.

marketing-ideas