auth

Pass

Audited by Gen Agent Trust Hub on Oct 12, 2026

Risk Level: SAFEDYNAMIC_CONTEXT_INJECTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [DYNAMIC_CONTEXT_INJECTION]: The skill uses the ! syntax in SKILL.md to guide users through interactive CLI tasks like mp login. This ensures that sensitive authentication flows (OAuth browser flows, service account secret entry) happen in the user's own terminal session rather than through the LLM context, which is a key security feature to prevent credential exposure.
  • [INDIRECT_PROMPT_INJECTION]: The skill manages data from external Mixpanel APIs, creating a potential surface for indirect prompt injection through malicious project or account names.
  • Ingestion points: Data enters the agent's context via the scripts/auth_manager.py helper, which utilizes the mixpanel_headless library to fetch project and account metadata.
  • Boundary markers: The instructions require the agent to parse JSON output and present it using specific visual structures (tables and single-line summaries), which provides clear boundaries for how external data should be processed.
  • Capability inventory: The skill uses a restricted Bash tool call to execute its management script, which has the capability to read configuration files and perform network requests to Mixpanel endpoints.
  • Sanitization: The auth_manager.py script acts as a sanitizer by mapping library objects to a strictly defined JSON schema before the data is interpreted by the agent.
  • [COMMAND_EXECUTION]: The skill uses a pre-approved Python script (auth_manager.py) to handle configuration tasks. The script uses the standard argparse library to handle inputs, mitigating command injection risks during script invocation.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 12, 2026, 03:23 PM