The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill identifies a sensitive credentials file path at ~/.brain_credentials.json which stores an email and password in plaintext. It includes specific instructions for the agent to avoid interacting with this file directly, advising the user to manage permissions manually via chmod 600.
[EXTERNAL_DOWNLOADS]: The scripts/wqo.sh script downloads the wqo==0.1.1 package from PyPI, a well-known service, when it is not found in the local environment. This follows standard dependency management practices.
[COMMAND_EXECUTION]: The skill uses a Bash runner to execute CLI commands and manage local virtual environments using uv or python, facilitating the operational capabilities of the WQO tool.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external tool output, including authentication states and biometric verification URLs. Ingestion points: CLI output processed in SKILL.md and references/runtime.md. Boundary markers: Explicit instructions to defer biometric verification to users and avoid reading credential files. Capability inventory: Subprocess execution and local file system access for environment setup. Sanitization: Relies on human-in-the-loop manual browser verification for external links.

wqo-auth