mx-core-local-auth
Warn
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: MEDIUMPRIVILEGE_ESCALATIONCREDENTIALS_UNSAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [PRIVILEGE_ESCALATION]: The skill directs the agent to bypass standard authentication gates by directly inserting a new session record with 'owner' privileges into the application's database, effectively bypassing intended login logic.
- [CREDENTIALS_UNSAFE]: The agent is required to access and utilize sensitive environment variables, specifically
PG_URLfor database connectivity andJWT_SECRETfor forging signed authentication cookies. - [COMMAND_EXECUTION]: Executing the instructions requires the agent to perform direct SQL database operations and HMAC-SHA256 cryptographic signing, which involves executing commands and logic on the host system.
- [INDIRECT_PROMPT_INJECTION]: The authentication bypass workflow is triggered by external stimuli such as API error codes or UI redirects, creating a surface where a malicious endpoint could potentially induce the agent to initiate high-privilege database operations.
- Ingestion points: UI and API response status codes (401, AUTH_NOT_LOGGED_IN) and redirect behaviors.
- Boundary markers: None present in the instructions.
- Capability inventory: SQL database writes, HMAC-SHA256 signing, and network request header modification.
- Sanitization: No validation or sanitization is specified for the triggering responses.
Audit Metadata