managing-experiment-lifecycle
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill serves as an operational guide for managing experiments within the PostHog platform, focusing on state transitions and administrative actions.
- [SAFE]: The instructions include explicit safety checkpoints, such as requiring the agent to confirm with the user before performing permanent actions like shipping a variant or opening a code cleanup pull request.
- [SAFE]: The 'Flag cleanup PR' feature is a documented platform capability that operates under specific permission scopes and requires user-initiated consent, reducing the risk of unauthorized code modifications.
- [SAFE]: No evidence of prompt injection, obfuscation, or unauthorized data exfiltration was found. The skill correctly handles errors by instructing the agent to explain failures to the user rather than attempting unsafe retries.
Audit Metadata