AIDA 1.1 ships auth, lighter pentest stack

The interesting part here is not just "LLM does pentesting" but that AIDA is closing the gap between agent demos and a usable security workbench. Swapping Exegol dependency for a built-in 2 GB container is the kind of practical product move that matters more than flashy autonomy claims.

• –The built-in container removes a major setup barrier, while still leaving Exegol as an optional heavier path for broader tool coverage
• –JWT auth, roles, first-run setup, and the Docker/socket hardening changes show the maintainer is reacting to real operator concerns, not just adding more agent tricks
• –`python_exec` gives the agent an escape hatch for edge-case payloads and workflow glue that fixed-tool wrappers usually miss
• –`http_request` plus credential placeholders makes the system more viable for authenticated web testing, which is where many autonomous security demos fall apart
• –The repo positions AIDA as model-agnostic and MCP-native, which makes it more interesting as infrastructure for security agents than as a single-purpose app