cdeust/zetetic-team-subagents

7 stars · Last commit 2026-09-29

One epistemic standard on three hosts. Claude Code: 15 problem-shaped skills, 97 sourced reasoning patterns, 120 agents, commit-time gates. Codex and Gemini CLI: portable evidence-synthesis, design and goal-loop skills. Shared gates block unsourced constants, vacuous tests and AI-writing patterns. I do not know is a supported outcome.

README preview

<p align="center">
  <img src="assets/banner.svg" alt="Zetetic reasoning for AI agents: sourced evidence, explicit uncertainty, one epistemic standard" width="100%"/>
</p>

<p align="center">
  <a href="https://github.com/cdeust/zetetic-team-subagents/actions/workflows/ci.yml"><img src="https://github.com/cdeust/zetetic-team-subagents/actions/workflows/ci.yml/badge.svg?branch=main" alt="CI"></a>
  <img src="https://img.shields.io/badge/suites-41-brightgreen" alt="Test suites">
  <img src="https://img.shields.io/badge/agents-120-8A2BE2" alt="Agents">
  <img src="https://img.shields.io/badge/skills-86-green" alt="Skills">
  <img src="https://img.shields.io/badge/hooks-19_lifecycle-red" alt="Hooks">
  <a href="LICENSE"><img src="https://img.shields.io/badge/License-MIT-blue.svg" alt="MIT License"></a>
  <a href="https://www.bestpractices.dev/projects/13847"><img src="https://www.bestpractices.dev/projects/13847/badge" alt="OpenSSF Best Practices"></a>
</p>

> **One epistemic standard on three hosts: portable skills for Codex and Gemini CLI, shared quality gates for Claude Code and Codex, and a full Claude-native distribution of 15 problem-shaped skills backed by 97 sourced reasoning patterns.**
> Three portable packages ship without the agent roster: evidence synthesis, UX design and accessibility audit, and goal-driven iteration (goal, plan, verify-goal, refine-goal, advisor-model). The mechanical gates run on both coding hosts from one shared runtime: source discipline and craftsmanship at commit, a fail-before check at push, secret-read refusal, deletion detection, and prose checks on returned and published text. The full Claude Code package additionally routes skills such as `causal-audit`, `failure-forensics`, and `estimation` across 97 genius agents (plus 23 team-role agents = 120 total), with the lifecycle hooks and the research and memory tooling.
> Not a prompt library. A methodology with **commit-time enforcement**.

---

View full repository on GitHub →