The Agent Skills Directory

[PROMPT_INJECTION]: No attempts to override agent behavior or bypass safety filters were detected. The instructions are strictly focused on copywriting principles and tone calibration.
[DATA_EXPOSURE]: The skill instructs the agent to read local context files like .agents/product-marketing.md to personalize emails. This is standard context-gathering behavior and does not involve unauthorized access or exfiltration. No hardcoded credentials or sensitive paths were found.
[REMOTE_CODE_EXECUTION]: There are no scripts, package installations, or remote code download patterns. The skill is entirely instructional markdown.
[OBFUSCATION]: The content is clear and uses standard Markdown formatting. No hidden strings, zero-width characters, or suspicious encodings were detected.
[INDIRECT_PROMPT_INJECTION]: While the skill processes user-supplied information about prospects to generate emails, it lacks any high-risk capabilities like shell execution, file writing, or network operations that could be exploited via injection. The risk remains low and consistent with standard LLM interaction patterns.
[DYNAMIC_CONTEXT_INJECTION]: The skill does not use shell command interpolation or dynamic context placeholders in its instructions.

cold-email