echo
Fail
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: HIGHPRIVILEGE_ESCALATIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [PRIVILEGE_ESCALATION]: The skill instructs the agent to execute the
Antigravity(agy) CLI tool using the--dangerously-skip-permissionsflag. This instruction explicitly directs the agent to circumvent standard security and permission protocols during its operations. - Evidence: Found in
reference/process-workflows.mdandreference/tri-engine-walkthrough.md. - [COMMAND_EXECUTION]: The skill utilizes external command-line interfaces for its 'Multi-Engine Mode,' including
codex exec --full-auto. The request for 'full auto' execution in a shell environment, combined with the aforementioned permission bypass, presents a significant risk for unintended or malicious system actions. - Evidence: Found in
reference/process-workflows.md. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, creating an attack surface for indirect prompt injection.
- Ingestion points: The skill ingests UI screenshots, mockups, and live URLs for evaluation (found in
reference/tri-engine-walkthrough.md). - Boundary markers: There are no explicit instructions or delimiters mentioned to segregate these external inputs from the agent's core instructions.
- Capability inventory: The skill has the capability to execute shell commands (
codex,agy) and spawn subagents. - Sanitization: No sanitization or validation logic is specified for the content extracted from images or URLs.
Recommendations
- AI detected serious security threats
Audit Metadata