The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or unauthorized data exfiltration behaviors were detected in the skill instructions or metadata.
[EXTERNAL_DOWNLOADS]: The skill provides a reference to an external repository (github.com/fabraix/playground) as a recommended resource for manual red-team validation. The skill does not automate any downloads or execution from this or any other external source.
[PROMPT_INJECTION]: The instructions include a dedicated safety notice that directs the agent to ignore any adversarial instructions found within the material being reviewed. This serves as a defensive hardening measure to prevent the analysis process from being manipulated by the content under review.

agent-security