whoami

Cross-AI user identity profile sync service. When you need user's background, preferences, or personal context to better execute a task, invoke this skill to...

Before installing or enabling this skill, consider: 1) The skill stores and uploads a personal profile to a remote service (default https://whoamiagent.com). Only use it if you trust that service and its privacy policy. 2) The skill requires an API key saved to ~/.whoamiagent (WHOAMI_API_KEY), but the registry metadata did not declare this credential — verify where that config comes from and inspect it before running. 3) The SKILL.md instructs using curl to download a one-time config URL and save it directly to your home config file; do not run those commands if you cannot verify the URL/source. 4) Environment variables (WHOAMI_ENDPOINT, WHOAMI_FRONTEND_URL) can redirect the skill to arbitrary endpoints — ensure these are unset or point to trusted hosts. 5) Avoid putting secrets or very sensitive personal data in the profile; the skill will POST profile content to a third-party API. 6) If you still want to try it, run the script in a controlled environment first (inspect ~/.whoamiagent after setup, test against a non-production endpoint, or sandbox it). If you want a safer alternative, request the skill author to add explicit registry metadata listing the required API key and to include an option to preview/validate any downloaded config before writing it to disk.

Type: OpenClaw Skill Name: whoami Version: 1.0.2 The `SKILL.md` file contains a critical prompt injection vulnerability. It explicitly instructs the AI agent to execute `curl -s "<provided-url>" > ~/.whoamiagent`, where `<provided-url>` is directly supplied by the user. This allows a malicious user to inject arbitrary configuration into the sensitive `~/.whoamiagent` file, including redirecting the `WHOAMI_ENDPOINT` to an attacker-controlled server. Consequently, all subsequent API requests made by `scripts/whoami_profile.py` would be directed to the malicious server, potentially leading to data exfiltration of user profile information or manipulation of the AI agent's operational context. While the Python script itself does not exhibit malicious intent, this vulnerability in the agent's instructions creates a significant attack vector.