The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest untrusted data such as scraped web text and emails. However, it contains explicit defensive instructions to treat these sources as evidence rather than instructions and warns the agent not to follow any embedded commands found within them. Additionally, the skill lacks any automated execution capabilities or scripts that could be leveraged by an injection.
[COMMAND_EXECUTION]: The skill instructions and reference materials emphasize that real-world side effects and persistent changes require explicit user authorization. No automated or hidden command execution patterns were found.
[CREDENTIALS_UNSAFE]: The reference materials include clear guidelines for developers to keep API keys out of skills and source control, following security best practices.

deliverability-qa