iblai-vibe-agent-sandbox

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

• Third-party content exposure detected (medium risk: 0.30). The required runtime workflow ingests outsider-authored free text via the user-facing “Prompts” editor in the Sandbox tab, which reads and upserts user-provided markdown content (IDENTITY.md/SOUL.md/etc.) into the agent’s runtime config before pushing to the connected OpenClaw/NemoClaw sandbox.

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

• Potentially malicious external URL detected (high risk: 0.90). The repo is explicitly cloned at runtime via "git clone https://github.com/iblai/vibe.git" (references/openclaw-install.md) and then installed into an OpenClaw workspace so that files from that external repo (notably AGENTS.md) are injected into the agent's system prompt, i.e., remote content fetched at install time directly controls agent prompts.