The Agent Skills Directory

[DATA_EXFILTRATION]: Sensitive authentication tokens and search queries are transmitted to an untrusted third-party domain.
The script scripts/search.py sends the user's WEB_SEARCH_BEARER token in an Authorization header to https://amd1.mooo.com/api/duck/search.
The use of mooo.com, which is a dynamic DNS provider, for a backend API is highly suspicious and constitutes a high risk of credential exfiltration.
[CREDENTIALS_UNSAFE]: The skill handles sensitive credentials in cleartext and transmits them to non-official endpoints.
Instructions in SKILL.md and code in scripts/search.py require users to provide tokens and SearXNG credentials (format: URL@USERNAME@PASSWORD) which are then processed and sent over the network.
[EXTERNAL_DOWNLOADS]: Dependencies are retrieved from unverified remote sources rather than standard registries.
The setup scripts install.sh and install.bat download and install the credgoo package directly from https://skale.dev/credgoo, bypassing the security controls of the PyPI registry.
[REMOTE_CODE_EXECUTION]: Installation scripts execute remote code fetched directly from the internet.
The install.sh script executes curl -LsSf https://astral.sh/uv/install.sh | sh and the install.bat script uses a similar PowerShell command. While Astral is a known developer, piping remote scripts to a shell is a dangerous pattern that executes arbitrary code.

web-search