The Agent Skills Directory

[SAFE]: The skill is composed of informational Markdown files and reference data. It does not contain any executable scripts, binaries, or command-line instructions.
[SAFE]: Explicit safety boundaries are defined in SKILL.md to prevent the agent from inventing facts, scraping prohibited data, or performing unauthorized actions such as sending emails or ignoring opt-outs.
[SAFE]: No instances of prompt injection, obfuscation, or persistence mechanisms were detected. All external URLs (GitHub and NVIDIA) are well-known or belong to the skill author and are used for documentation purposes.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect injection as it processes user-provided prospect data to generate email drafts.
Ingestion points: User prompts containing prospect roles, company names, and research signals.
Boundary markers: Present in SKILL.md via the "Boundaries" and "Quality Check" sections which restrict the agent's behavior.
Capability inventory: None. The skill does not possess tools for file system modification, network access, or command execution.
Sanitization: Not explicitly implemented for inputs, but the lack of functional capabilities ensures that any potential injection remains confined to text generation without system impact.

suede-cold-email