axiom-axe-ref
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill provides installation instructions
brew install cameroncooke/axe/axe, which downloads the AXe CLI binary from a third-party repository that is not part of the trusted vendor list. - [COMMAND_EXECUTION]: The reference includes numerous shell commands for the
axetool to perform UI automation tasks such as tapping elements, typing text, and simulating hardware button presses on an iOS simulator. - [INDIRECT_PROMPT_INJECTION]: The skill uses
axe describe-uito inspect simulator UI hierarchies, which creates a surface for indirect prompt injection if the agent processes untrusted element labels or identifiers. * Ingestion points: UI accessibility tree data fromaxe describe-uimentioned inSKILL.md. * Boundary markers: Not specified in the instructions. * Capability inventory: UI interaction (tap, swipe, type), hardware button simulation, and local file writing (screenshots, video). * Sanitization: No sanitization or validation of the ingested UI content is described.
Audit Metadata