The Agent Skills Directory

[SAFE]: The skill consists entirely of instructional markdown content, templates, and marketing strategies. It does not include executable scripts, shell commands, or package installations.- [DATA_EXFILTRATION]: The skill recommends reading local context files like .agents/product-marketing.md. This is a routine operation for gaining project context and does not involve hardcoded credentials or unauthorized data transmission to external servers.- [PROMPT_INJECTION]: No patterns were detected that attempt to override system instructions, bypass safety filters, or extract system prompts.- [EXTERNAL_DOWNLOADS]: The skill mentions external tools like Mailchimp and SendGrid in a reference table, but these are links to internal documentation and do not trigger remote downloads or code execution.- [INDIRECT_PROMPT_INJECTION]: The skill includes instructions to process external context data which represents a theoretical attack surface.
Ingestion points: Reads .agents/product-marketing.md, .claude/product-marketing.md, and product-marketing-context.md as described in SKILL.md.
Boundary markers: Absent.
Capability inventory: The skill is limited to text generation and lacks capabilities for command execution, network requests, or file writing.
Sanitization: Absent.
Assessment: The risk is safe because the agent lacks the necessary tools to be exploited by malicious content within the context files.

emails