mimiry-softlaunch

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly fetches and ingests content from the Mimiry public API (e.g., GET /api/compute/v1/availability and GET /api/compute/v1/sessions//logs as shown in SKILL.md and scripts/mirc.sh), and those responses — including user/container-generated session logs and availability/provider data — are parsed and used to make decisions (pick GPU/provider, wait for readiness, construct SSH/creation commands), so untrusted third-party content can materially influence agent actions.