How abhinav16aero/purpleteam works
RedBlue AI coordinates Decepticon red-team engagements with Vigil blue-SOC detection, then scores and governs the response. Users initiate engagements through the coordinator API; the coordinator triggers attacks, collects detections, computes defensive posture, and produces a scorecard and evidence record. The README documents the engine and sensor flow; source excerpts confirm the coordinator’s continuous-mode and policy components. Other coordinator stages and integrations are represented at documented directory-level paths because their implementation was not sampled.
Main entry points: coverage.py Coverage scoring · agents Red agents · vigil Vigil SOC · app.py Engagement API
Components 18
Coordinator
- Engagement API
redblue-coordinator/redblue/api/app.py - Engagement planning
redblue-coordinator/redblue - Red trigger
redblue-coordinator/redblue - Telemetry collection
redblue-coordinator/redblue - Coverage scoring
redblue-coordinator/redblue/scoring/coverage.py - Continuous replay
redblue-coordinator/redblue/continuous/controller.py
Red team
- Red agents
Decepticon/packages/decepticon/decepticon/agents - Injection shield
vigil-llm/vigil/core
Blue team
- Vigil SOC
vigil - Injection shield
vigil-llm/vigil/core
Governance
- Policy engine
redblue-coordinator/redblue/governance/policy_engine.py - Governed response
redblue-coordinator/redblue - Engagement scorecard
redblue-coordinator/redblue/scoring - Evidence ledger
redblue-coordinator/redblue/evidence
Other components
- Operator
- Target range
- Security sensors
- Local Ollama
Connections 18
- Operator → Engagement API: launches
README.md - Engagement API → Engagement planning: starts
README.md - Engagement planning → Red trigger: triggers
README.md - Red trigger → Red agents: launches
README.md - Red agents → Target range: attacks
README.md - Target range → Security sensors: observed by
README.md - Security sensors → Vigil SOC: sends telemetry
README.md - Vigil SOC → Telemetry collection: returns findings
README.md - Telemetry collection → Coverage scoring: provides detections
README.md - Coverage scoring → Engagement scorecard: produces
README.md - Coverage scoring → Policy engine: informs decisions
- Policy engine → Governed response: governs
README.md - Governed response → Evidence ledger: records decision
README.md - Continuous replay → Evidence ledger: writes evidence
redblue-coordinator/redblue/continuous/controller.py - Red agents → Injection shield: uses shield
README.md - Vigil SOC → Injection shield: uses shield
README.md - Injection shield → Local Ollama: local inference
README.md - Injection shield → Local Ollama: local inference
README.md