Enables agents to probe URLs and classify bot-wall responses with vendor and signal evidence, perform offline classification, scan for exposed key-like material, and run owner-side access audits through stdio tools.
Enables agents to analyze software artifacts through 35 tools covering read-only reconnaissance, secret/string/symbol scanning, native and Dart AOT structure checks, exposure tiering, protection planning, attack-simulation design, and an append-only evidence ledger with mechanical verified gates that treat unresolved unknowns as blocking. Gated execution tools apply reviewed transforms only to a managed copy inside an authorized workspace, with one-time approval tokens, independent verification, before/after diffs, and rollback.
It lets an AI assistant call on a separate configured model for bounded, one-at-a-time adversarial checks — pressure-testing early ideas and claims, tracing failure paths, examining documents for structural weaknesses, exploring abuse paths, and proposing experiments — across planning, product, procedure, technical, and security work. The calling assistant stays in charge, deciding which returned criticisms the supplied material actually supports and discarding the rest.
Enables auditing and monitoring of AI agent MCP tool calls, with behavioral anomaly detection, lab-scoped red-team simulation, and Telegram notifications.
A vulnerable-by-design MCP server pair (NotesServer and VaultServer) for testing MCP security tools, featuring confused-deputy, prompt injection, and authorization bypass scenarios.