3 The taxonomy
The Kailash catalogue is organised into four named tool layers (plus a CORE substrate that hosts none of them), and within each layer into numbered categories. The catalogue of record is the tool manifest — a single hand-maintained registry in the packages repository; packaging, menus, CLI and the generated docs all derive from it.
Tool counts below are the census at this guide’s build (337 rows across 32 categories; the manifest always carries the live numbers).
3.1 Layer CLASSIC (c-01 … c-10) — 147 tools
Software-only heritage security foundation: information gathering (38), vulnerability analysis (7), web application assessment (10), database assessment (7), password & offline-credential attacks (16), reverse engineering (26), Signal & SDR — the opt-in carve-out (6), exploitation & post-exploitation, software-only (9), forensics/IR & reporting (21), runtime AppSec for AI systems (7).
3.2 Layer ATTACK (a-01 … a-08) — 58 tools
Offensive AI: AI reconnaissance & target discovery (5), LLM prompt injection & jailbreaking (8), agentic/MCP/tool-use attacks (9), model supply chain & artefact poisoning (8), model extraction/inversion & membership inference (8), multi-modal & physical-AI attacks (5), AI system exploitation (9), data poisoning & dataset attacks (6).
3.3 Layer DEFENCE (d-01 … d-07) — 45 tools
Defensive AI & assurance: runtime defence & guardrails (7), detection/ monitoring/observability & IR (8), assurance/evals/benchmarks (6), model provenance/signing/trust (10), AI threat modelling (6), governance & regulation mapping (4), standards verification (4).
3.4 Layer OPS (o-01 … o-07) — 87 tools
The platform: MLOps & ML engineering (18), compute & serving infrastructure (11), vector stores & retrieval (9), local LLM & inference labs (5), AI engineering SDKs & dev environment (26), documentation/reporting/authoring (10), agent & MCP operability (8).