The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or unauthorized access attempts were identified. The skill defines a controlled autonomous workflow with explicit human oversight and security boundaries.
[SAFE]: The 'authorization envelope' feature explicitly restricts dangerous side effects, such as external data exfiltration or unauthorized spending, by default.
[SAFE]: The framework enforces 'identity checks' and state verification to prevent context confusion or unauthorized resumption of tasks.
[SAFE]: The generated report.html is mandated to be self-contained and prohibited from making external network calls, mitigating risks of data exfiltration via the dashboard.

long-horizon