integrations-sh

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

• Third-party content exposure detected (low risk: 0.10). In cli.ts, outsider free text provided by the user can be sent into integrations.sh’s agent-facing workflow via runtime query parameters (e.g., integrations search <query> / domain term args are used to build the request URL and then the CLI prints returned text), but the required workflow itself does not ingest arbitrary third-party content unless it is fetched from the integrations.sh API based on those arguments.

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

• Potentially malicious external URL detected (high risk: 0.90). The CLI fetches the OpenAPI spec and registry from the integrations.sh host at runtime (BASE defaults to "https://integrations.sh", used to request /openapi.json and /api.json), and that fetched spec directly controls the command surface and help text, so remote content can alter agent prompts/behavior.