partiful

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). The code deliberately captures bearer auth tokens and user IDs by intercepting browser network traffic (via a Playwright script it writes and executes), stores them locally, and also auto-installs/runs external tooling (npm/playwright), which together represent intentional credential-capture and execution of external code that can be abused or repurposed for exfiltration or supply-chain compromise.

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

• Third-party content exposure detected (high risk: 0.85). The skill fetches outsider-authored free text at runtime from Partiful’s public web/API responses (e.g., cmd-invites → get-my-rsvps → api-request → JSON fields like event titles/descriptions), which is then printed and included in the agent’s LLM context via the tool output.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill's auth command installs and runs Playwright at runtime (it executes "npm install -g playwright" / "npx playwright install chromium" and then runs a Node script), which fetches and executes remote code from the npm registry (e.g. https://registry.npmjs.org/playwright), so the skill fetches and executes external code during runtime.