The Agent Skills Directory

[SAFE]: A detailed security audit of the main skill file and all reference materials confirms that the skill is safe and operates within its intended scope. No findings related to prompt injection, data exfiltration, or credential exposure were identified.
[INDIRECT_PROMPT_INJECTION]: The skill instructions specify reading external marketing context from files such as .agents/product-marketing.md. Ingestion points: Reads content from local project files (SKILL.md). Boundary markers: The instructions do not define specific delimiters for separating file content from internal logic. Capability inventory: None; the skill's logic is limited to text generation and does not utilize tools for network access, file writing, or system command execution. Sanitization: No explicit sanitization of the context file content is performed. Conclusion: This ingestion point is a standard feature for context-aware agents and poses no significant risk due to the lack of exploitable capabilities within the skill.

copywriting