iblai-api-agent-embed
Pass
Audited by Gen Agent Trust Hub on Sep 10, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill facilitates the management of agent settings and user authentication by performing authenticated network requests to the
iblai.appdomain. These operations include high-privilege tasks such as minting security tokens (axd_token,dm_token,edx_jwt_token) and configuring agent behavior. The skill mitigates risk by requiring explicit user confirmation before executing these sensitive API calls. - [INDIRECT_PROMPT_INJECTION]: The skill serves as a management interface for processing user-supplied data that influences agent appearance and authentication.
- Ingestion points: External data enters through user-provided parameters for the API payloads, including user identity information (username, email) and custom scripts (
custom_css,custom_javascript). - Boundary markers: The skill uses explicit instructional markers, requiring the agent to "Confirm with the user first" before initiating sensitive operations like share link generation, token redirection, or user provisioning.
- Capability inventory: The skill utilizes authenticated HTTP requests (PUT/POST) via the platform API to modify agent embed settings and provision user credentials.
- Sanitization: No specific client-side sanitization is described for the
custom_javascriptorcustom_cssfields; the skill relies on the platform's backend validation and the agent's adherence to safety protocols during the confirmation step.
Audit Metadata