grillkit
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill defines a complex instructional framework for design interrogation. It uses standard platform tools like Bash, Write, and Edit for project documentation management and Git integration without any suspicious patterns detected.
- [COMMAND_EXECUTION]: The skill uses shell commands for repository management (
git add,git commit,git push) and issue tracking (gh issue edit). These are standard capabilities for development-oriented agents. The skill mitigates risks by explicitly requiring a confirmation step ("Publish the result") where the user must approve these actions before they are executed. - [INDIRECT_PROMPT_INJECTION]: As the skill processes user-provided ideas and plan files, it technically has an ingestion surface for indirect prompt injection. However, the core logic requires the agent to reflect ideas back to the user and iterate through structured rounds of questioning. This multi-step, human-reviewed process serves as an effective validation layer against adversarial instructions embedded in data.
- [DATA_EXPOSURE_AND_EXFILTRATION]: No attempts to access sensitive system files (e.g.,
.env, SSH keys, cloud credentials) or exfiltrate data to untrusted remote servers were identified. All network-related operations are scoped to the user's configured Git and GitHub environment.
Audit Metadata