kadath

name	kadath
description	Run KADATH (Kernel for Agentic Darwinian Adaptation, Tooling, and Heredity), a Docker-based evolutionary kernel that turns a goal into a locked, Architect-authored benchmark, then evolves a population of smolagents-based coding agents across epochs: each agent runs in an isolated container, gets graded against frozen evidence, and the population is culled, mutated, and reproduced generation over generation until it converges on the best-performing agent framework for that goal. Use when the user wants to propose/approve/run a KADATH evolutionary run, check a run's status or live dashboard, pause/resume/continue a run, export the winning agent population, or understand its Architect/Grader/Tweaker/Birther pipeline, evidence-freezing, or genome lineage/memory model. Triggers on: "kadath", "kadath.sh", "evolve an agent", "Darwinian agent evolution", "agent population fitness benchmark", "smolagents evolutionary run", "kadath dashboard", "genome lineage", "epoch champions".
allowed-tools	Bash Read Write Edit Glob Grep
compatibility	Requires Docker Engine plus the Docker Compose plugin, an OpenAI API key, and a host with a real TTY for the interactive `./kadath.sh` frontend (non-interactive automation uses the `kadath` CLI directly). Python 3.11+ backs the `kadath` package declared in `pyproject.toml`. Apache-2.0 license.
metadata	{"tags":"kadath, evolutionary-agents, smolagents, multi-agent, agent-benchmarking, docker, llm-agents, genetic-algorithm, python","platforms":"Claude, ChatGPT, Gemini, Codex","version":"1.0","source":"https://github.com/i3T4AN/KADATH"}

KADATH — evolutionary agent kernel

KADATH takes a goal and spends model tokens evolving agents that get progressively better at achieving it. A population of smolagents-based CodeAgent organisms competes each epoch, gets independently graded against a locked, Architect-authored benchmark, and is then culled, reflected on, mutated, and reproduced — generation over generation — under a kernel that owns the run, containers, evidence freezing, grading formulas, and Git-backed genome lineage. The organisms are read-only while an epoch runs and can only change during the post-grade mutation phase, so improvement happens through repeated competition and selection rather than one prompt or one agent.

When to use this skill

• Standing up a new KADATH run: proposing a goal, letting the Architect draft a benchmark, and approving it (kadath init/kadath start/kadath approve)
• Launching, pausing, resuming, or continuing an evolutionary run, or watching its live dashboard
• Exporting a finished run's winning agent population, or continuing evolution from one specific historical genome
• Explaining or debugging KADATH's Architect/Grader/Tweaker/Birther pipeline, evidence-freezing, container isolation, or memory/heredity model to a user working in this codebase
• Running the read-only local Docker stack (kadath.sh) that provides PostgreSQL, MinIO, LiteLLM, SearXNG, and Playwright MCP for a run

When not to use this skill

• Building or fine-tuning a single agent by hand with no evolutionary/competitive-selection element → use a normal agent-framework or fine-tuning skill instead
• Generic multi-agent orchestration without grading, culling, and reproduction across generations → KADATH's whole value is the selection loop, not just running agents in parallel
• The user wants a lightweight, no-Docker local script → KADATH's control plane requires Docker Compose (PostgreSQL, MinIO, LiteLLM) and is not designed to run bare

Instructions

Step 1: Clone and read the operational contract first

git clone https://github.com/i3T4AN/KADATH.git
cd KADATH

Read README.md fully before running anything — it documents the two-layer design (kernel vs. organisms), the Architect's machine-readable benchmark contract, isolation/credential rules, and recovery behavior. kadath/engine.py is the run state machine; kadath/cli.py is the direct CLI surface; seed/organism.py is the default evolvable agent loop.