nadiwerk/trust-no-agent

{
  "$schema": "https://agent-plugins.org/schemas/1.0.0/plugin.schema.json",
  "name": "trust-no-agent",
  "version": "0.1.26",
  "description": "Trust over speed for coding agents — the router (AGENTS.md + WORKFLOW.md), 11 discipline skills, and the mechanical enforcement layer: floor re-injection and per-prompt MANDATORY classification.",
  "author": {
    "name": "nadiwerk"
  },
  "homepage": "https://github.com/nadiwerk/trust-no-agent",
  "repository": "https://github.com/nadiwerk/trust-no-agent",
  "license": "MIT",
  "keywords": [
    "codex",
    "codex-plugin",
    "hooks",
    "skills",
    "discipline",
    "agents-md",
    "trust",
    "verification"
  ],
  "skills": "./skills/",
  "hooks": "./hooks/codex.hooks.json",
  "extensions": {
    "com.openai": {
      "interface": {
        "displayName": "Trust No Agent",
        "shortDescription": "Discipline the agent will not skip",
        "longDescription": "Four Iron Laws, eleven discipline skills, and a mechanical boundary the model does not have to remember: the session-start floor is re-injected from the router, and every prompt is classified against the MANDATORY domains (expect-fail, root-cause, receipts) with a LOAD directive injected when one is touched. Self-trigger measured 0/3 — the hooks exist because relying on the model to remember fails silently.",
        "developerName": "nadiwerk",
        "category": "Developer Tools"
      }
    }
  }
}