linear

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (low risk: 0.10). The Linear skill’s runtime workflow interacts with Linear via its MCP server/GraphQL tool, where the model can read issue/comment/document “body/title/content” fields from the user’s Linear workspace, but the skill itself does not describe any outsider-authored free-text submission path that is automatically ingested without the user selecting/fetching specific items.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The agent manifest declares a required MCP tool with the runtime URL https://mcp.linear.app/mcp (agents/openai.yaml:14), which the agent will contact at runtime and which can execute remote tool logic on behalf of the agent.