1 · What Strix is
Strix spins up a disposable Docker sandbox loaded with real pentest tools, then hands control to a tree of AI agents — one orchestrator ("root"), many specialists ("children") — that actually probe a target, validate findings with a working proof-of-concept, and write a CI-ready report.
strix.core.runner.run_strix_scan() is the single entry
point every mode (headless CLI, interactive TUI) calls. It builds one
AgentCoordinator, one shared Docker sandbox, a root
SandboxAgent, and awaits core.execution.run_agent_loop() for
it — child agents are spawned as independent asyncio.Tasks sharing the
same sandbox.2 · Who should read what
| Reader | Start with | Then |
|---|---|---|
| Non-technical | §1 above | §3 Module Map (Simple mode) |
| Security analyst | §3 Module Map | docs/design/data-flow.md |
| Junior developer | §3 Module Map (Technical) | docs/modules/*.md |
| Reviewer / architect | docs/design/HLD.md | docs/design/LLD.md |
3 · Module map — click any box
Every box answers four questions: what happens, inputs/outputs, how the model is chosen (or "no model calls"), and what's deterministic vs. AI-driven.
Interface
CLI, interactive Go TUI, and the local read-only web viewer (strix view).
Core Orchestration
The multi-agent orchestration engine: AgentCoordinator, the turn loop, resume.
Agent Construction
Assembles one agent's tools and renders its system prompt.
Agent Tools
Everything an agent can call: shell/browser/proxy, agent-graph, reporting, notes.
Runtime (Sandbox)
Creates the one shared Docker sandbox per scan and its Caido intercepting proxy.
Findings
The in-memory + on-disk record of every finding, deduplicated and written out.
Config & Model Routing
Resolves the LLM model/provider and every other tunable, once per scan.
Internal Skills (Knowledge Packs)
Markdown methodology library (vulnerabilities, tooling, coordination, scan modes).
4 · The Graph of Agents
One "lead" agent never tests anything itself — it only reads the brief and assigns specialists. Each specialist can itself assign sub-specialists for a sub-task. When a specialist is done, it mails a summary back to whoever assigned it.
AgentCoordinator (strix/core/agents.py) is the sole
owner of this tree: parent_of tracks structure, send()/mailboxes
carry messages, and the whole state is periodically snapshotted to disk so
strix --resume can reconstruct it after a crash. See
diagrams/05_agent_graph_orchestration.svg.
5 · Explore the full documentation
Every deliverable this kit produced, one click away. Markdown is always the source of truth for anything with an .html mirror below — regenerate the HTML, never hand-edit it.
Diagrams
| Diagram | Static (svg/png) | Interactive |
|---|---|---|
| Architecture (layered view) | 01_architecture.svg | 01_architecture.html |
| Data flow (one scan, start to finish) | 02_dataflow_pipeline.svg | 02_dataflow_pipeline.html |
| Context diagram (DFD Level 0) | 03_context_dfd.svg | 03_context_dfd.html |
| Module dependencies (real CodeGraph data) | 04_module_dependencies.svg | — |
| Agent-graph orchestration | 05_agent_graph_orchestration.svg | — |
| Docker sandbox contents | 06_sandbox_container.svg | 06_sandbox_container.html |
| Standalone module map | — | diagrams/module-graph.html |
Design docs
| Doc | Markdown | HTML |
|---|---|---|
| High-Level Design | HLD.md | HLD.html |
| Low-Level Design | LLD.md | LLD.html |
| Data flow (DFD 0/1/2) | data-flow.md | — |
| Agentic framework & deterministic vs. LLM tooling | agentic-framework.md | agentic-framework.html |
Module docs (all 12 packages)
| Package | Markdown | HTML |
|---|---|---|
| agents | agents.md | agents.html |
| core | core.md | — |
| tools | tools.md | — |
| runtime | runtime.md | runtime.html |
| report | report.md | report.html |
| interface | interface.md | — |
| config | config.md | — |
| skills | skills.md | — |
| llm | llm.md | — |
| telemetry | telemetry.md | — |
| utils | utils.md | — |
| strix (root) | 00_root.md | — |
Full index with real file/LOC counts: docs/modules/README.md
Process & runbooks
| Doc | Markdown |
|---|---|
| Index | 00_process-and-runbooks-index.md |
| SOP | process-operate-strix.md |
| Process optimization | process-optimization.md |
| Runbook: setup & credentials | setup-and-credentials.md |
| Runbook: run a detection scan ⚠ | run-a-detection-scan.md |
| Runbook: review & triage findings | review-and-triage-findings.md |
Real code-intelligence tool output
| View | Open |
|---|---|
| codeviz — module → file → function drill-down | strix-codeviz.html |
| Stage explorer — S0–S15 → files → functions | strix-stage-explorer.html |
| Graphify — named clusters (197 communities, no LLM) | labeled-clusters.html |
| Graphify — curated call-flow diagrams | strix-callflow.html |
| Graphify — collapsible tree explorer | GRAPH_TREE.html |
| Graphify — raw force-directed graph (dense) | graph.html |
Full index and how these were produced: docs/tools/graph-intelligence.md, tool-outputs/README.md
Start here / overview
| Doc | Markdown |
|---|---|
| Start here (persona reading paths) | docs/00_START_HERE.md |
| Overview | docs/01_overview.md |