coordination

You coordinate autonomous pentest engagements by mounting the coordination Skill, which spawns executor and validator agents with context-aware prompts to maintain disciplined progress across the full engagement lifecycle.

Orchestration maintains attack-chain context by keeping a living attack-chain.md file that records hypotheses, iterative experiments, and results to guide ongoing testing and phase transitions.

Automating pentest validation and reporting workflows is handled by spawning validator agents and skeptic checkpoints that enforce governance rules, log provenance, and enable phase-3 handoffs to update skills and notify stakeholders.

You can run autonomous pentest orchestration without external dependencies by defining OUTPUT_DIR and TARGET variables, letting the centralized orchestration layer manage the engagement state internally.

Phase transitions between reconnaissance, experiments, and validation are managed by the orchestration layer, which uses centralized context and governance rules to trigger route-appropriate agent spawning.

Centralized orchestration is not suited for ad-hoc, single-step tests, as it enforces structured phase transitions, provenance logging, and iterative experimentation rows designed for full engagement lifecycles.