HydraSight

AI-assisted, authorization-gated offensive-security orchestration.

HydraSight drives a Kali backend (via the kali-server-mcp REST bridge) and a local LLM (Ollama) to plan, execute, verify, and report on authorized penetration tests. It is built around three non-negotiable safety properties:

1. Deny by default. No command reaches a target until an operator has explicitly attested authorization for that scope.
2. A single enforcement chokepoint. Every command is validated, scope-checked against both the Rules of Engagement and the operator attestation, and audited in one place (Dispatcher).
3. A tamper-evident audit trail. Every allow/block decision is written to an append-only, SHA-256 hash-chained JSONL log that can be verified at any time.

⚠️ Authorized testing only. HydraSight is intended for systems you own or have explicit written permission to test. Unauthorized use is illegal.

Why HydraSight?

• LLM orchestration with guardrails — plain English never executes tools; only explicit /run actions and confirmed plans dispatch, and every tool call is validated by a strict command sanitizer.
• Accountability — mandatory authorization attestation plus a hash-chained audit log with automatic secret redaction.
• Reporting — JSON engagement exports and PDF reports out of the box.
• Installable — a real PyPI package, a non-root Docker image, and a docker-compose stack with a Kali bridge.

Quick links

• Getting Started
• Security Model
• Authorization & Audit
• Rules of Engagement
• Usage

License

Released under the MIT License.