skills/getsentry/sentry/seer-embed/Gen Agent Trust Hub

seer-embed

Pass

Audited by Gen Agent Trust Hub on Sep 13, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill guides the agent to execute a local development script pnpm gen:embed-widgets. This script is used to synchronize the frontend Zod schemas with the backend JSON schema, which is a standard procedure in the project's development lifecycle.
  • [INDIRECT_PROMPT_INJECTION]: The skill describes a system for rendering UI widgets based on data generated by an AI agent. This establishes a potential indirect prompt injection surface where the agent's output is processed by the frontend. However, the framework incorporates strong mitigations: 1) Ingestion points: Untrusted JSON data is embedded within markdown using specialized tag syntax. 2) Boundary markers: The Markdoc-style tag syntax clearly delimits the data segment from the surrounding prose. 3) Capability inventory: The resulting React components are primarily used for data visualization and navigation (e.g., links, cards). 4) Sanitization: The defineSeerEmbed wrapper mandates the use of zod.safeParse to validate all incoming data against a predefined schema before rendering.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 13, 2026, 06:58 AM
Security Audit — agent-trust-hub — seer-embed