What problem does it solve? When running an authorized security engagement, choosing which attack-surface skills to load for a given target is error-prone: modern apps return soft-404s that fake findings, hosts run mixed stacks, and loading every matching skill blows the context window. This Skill acts as the dispatcher for the /hunt orchestrator, turning a mode answer (redteam or wapt, blackbox or greybox) into a deterministic, budget-capped skill loadout. ## Core Features & Use Cases - Soft-404 baseline: Records a per-host control (status, byte length, body hash) from bogus paths so no path counts as found unless its response differs from the control. - Fingerprint-driven routing: Maps banner and body markers (Okta, Entra, Next.js, Laravel, Spring Boot, gRPC, Kubernetes, and more) to platform and hunt-* skills, with tier-based priority, an 8-skill load cap, and de-duplication rules. - Mode-specific loadouts and taxonomy print: Loads always-on, platform, and hunt-* sets for redteam or wapt modes, validates greybox credentials with an auth preflight, and emits a deterministic taxonomy block before returning control to /hunt. - Use Case: After confirming an engagement is a greybox WAPT, invoke the dispatcher to baseline each host, validate the supplied session cookie against an identity endpoint, load the full OWASP-relevant hunt-* set, and print the loaded-skills taxonomy. ## Quick Start Run hunt-dispatch with mode=wapt box=greybox after /hunt collects the engagement mode and credentials.
