The Agent Skills Directory

[SAFE]: The skill provides clear instructions on handling untrusted external data, specifically warning the agent to treat scraped text and reviews as evidence rather than instructions to prevent indirect prompt injection.\n- [SAFE]: Secret management guidelines are included in the connector documentation, advising that API keys be kept out of skill files and source control.\n- [SAFE]: The skill lacks executable code or scripts, reducing the risk of automated malicious actions. It relies on a contract-based approach for permissions and side effects.

analytics-tracking