figure-it-out
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides a purely instructional framework for managing complex tasks through defined phases (Frame, Design, Run, Audit, Verify). It advocates for security and engineering best practices, such as using falsifiable predicates and verification harnesses.
- [SAFE]: The configuration in
agents/openai.yamldisables implicit invocation (allow_implicit_invocation: false), which is a security-hardening measure that limits how the agent can be triggered. - [SAFE]: The
SKILL.mdfrontmatter includesdisable-model-invocation: true, further restricting the execution environment and reducing the potential attack surface. - [SAFE]: No indicators of prompt injection, data exfiltration, obfuscation, or unauthorized command execution were found. The skill relies on modular, internal 'principle skills' to perform its tasks, promoting a structured and auditable environment.
Audit Metadata