User Authentication System

Role-based access control for Greek accounting firms. Login, role hierarchy, per-client permissions, session management, audit logging.

This skill looks like it does what it says (a local RBAC/auth system) but contains cryptographic and documentation inconsistencies that matter for security. Before installing: 1) Do not accept SHA‑256 password hashing—require the author to use a modern password KDF (bcrypt, scrypt, or Argon2) with explicit parameters. 2) Ask for details on how salts and TOTP secrets are generated, stored, and protected (secrets should be encrypted at rest and access-limited). 3) Verify that audit logs don't leak credentials or tokens and that file permissions restrict access to the auth directory (chmod 700 is suggested, but confirm ownership). 4) Confirm the discrepancy between SKILL.md (SHA‑256) and the EVALS tests (bcrypt) is resolved — this could be a documentation error or an implementation bug. 5) Prefer a non-default OPENCLAW_DATA_DIR (avoid using a global /data if that is shared) and review the actual implementation source before trusting it with production accounts. If you cannot obtain satisfactory answers or the owner/repo provenance, treat this as unsafe to deploy for real user accounts.

Type: OpenClaw Skill Name: user-authentication-system Version: 0.1.1 The OpenClaw AgentSkills bundle for 'user-authentication-system' is classified as benign. The skill is designed as a robust, local authentication and authorization system, employing strong security practices such as bcrypt for password hashing, secure session token management, 2FA, account lockout, and comprehensive audit logging. Crucially, the `SKILL.md` documentation explicitly outlines critical file system permission hardening (`chmod 700 /data/auth/`, `chmod 600 /data/auth/users/*/credentials.json`) required to protect sensitive data, demonstrating an awareness of security implications within the OpenClaw environment rather than malicious intent. There is no evidence of data exfiltration, backdoors, or prompt injection attempts against the agent in the provided content; the `SKILL.md` serves as legitimate user instructions.