Ouroboros

A plugin that puts Claude Code through an interview on a vague request: spec and code generation is blocked until the brief scores clearly enough

Plugin

Medium risk

We rate an entry medium when the tool runs code, makes network calls or reads project files. Check what exactly it does before installing.

Why this level

  • It runs an MCP server and executes generated code as part of the development loop
All reasons and checks

q00/ouroboros

Install

In your terminal, with SkillFoxx CLI

npx skillfoxx add plugins/ouroboros

Detects the agents on your machine, checks the risk and pins the version.

Other ways to install

Run one by one in the Claude Code chat

/plugin marketplace add q00/ouroboros
/plugin install ouroboros@ouroboros

Checked against the repository on Sep 25, 2026, commit a7e92a3.

Text for your agent

Install the plugin: /plugin marketplace add Q00/ouroboros, then /plugin install ouroboros@ouroboros. The package is also available via pip install ouroboros-ai.

Other ways from the author
/plugin marketplace add Q00/ouroboros
/plugin install ouroboros@ouroboros

This is third-party code. Review the repository files before installing.

What it does

Ouroboros sits in front of Claude Code's normal work and first scores how vague the task request is. Above a threshold, the agent will not move to spec and code generation until a short clarifying interview passes, or the user explicitly forces a skip with the force flag. The task then goes through acceptance-criteria setup, execution and verification, and the worker is handed the goal itself rather than the verification command's exact assertion code, so the task doesn't turn into fitting the test. It works as a Claude Code plugin with its own MCP server and skill set, and supports 14 runtimes including Codex CLI, Gemini CLI, OpenCode and Copilot.

Who it is for. For developers who want a safeguard against the agent writing code from a vague request and missing the actual goal.

Good fit when

  • The task is stated briefly and loosely, but the result needs to be predictable
  • You need an explicit acceptance-criteria step before the agent starts writing code
  • Work spans several agent runtimes and needs one shared task-framing process

Not a fit when

  • The task is already stated with full clarity, and an interview would just slow things down
  • You need a quick one-off run without extra questions from the agent

Example request

Run this feature request through ouroboros and interview me before writing the spec

Limitations

Redaction of verify-command values and expected-output assertions elsewhere in the contract is a literal match across five encodings, so a reshaped copy of a secret can still slip through; this is a known, publicly tracked gap. It requires installing the plugin and the ouroboros-ai Python package.

How to disable. Remove the plugin with /plugin uninstall ouroboros or uninstall the package with pip uninstall ouroboros-ai.

Security check

  • It runs an MCP server and executes generated code as part of the development loop

README in short

The README calls Ouroboros an Agent OS for replayable coding iterations: an interview gated on request vagueness, staged evaluation, and a budgeted evolution loop. It lists an ouroboros-ai PyPI package, support for 14 runtimes, and its own MCP server. The repo has a .claude-plugin with marketplace.json and plugin.json, a .mcp.json for server setup, and a skills folder with a separate SKILL.md per stage, including auto, brownfield, cancel and config.

FAQ

Can the interview be skipped?

Yes, with an explicit force flag if the user is confident the brief is sufficient even though it scores below the bar.

Does it work beyond Claude Code?

Yes, the README lists 14 supported runtimes, including Codex CLI, Gemini CLI, OpenCode, Copilot and Kiro.

Editors’ pick

A skills library that gives coding agents a development process: brainstorming, planning, TDD, subagents and code review

PluginMedium riskNo VPN needed292.5KRepository stars
Editors’ pick

Small composable skills for engineering with agents: plan grilling, TDD, bug diagnosis, code review and architecture

SkillLow risk271.4KRepository stars
Editors’ pick

GitHub toolkit for spec-driven development: the specify CLI adds agent commands and skills to a project, from principles to implementation

CLIMedium riskNo VPN needed139.3KRepository stars

Reference MCP servers

Model Context Protocol servers

Official

Official reference MCP servers: Filesystem, Fetch, Git, Memory, Sequential Thinking, Time and Everything

MCP serverMedium risk90.6KRepository stars
Foxx AIOuroboros

I am Foxx AI and I have already vetted this tool. Ask about install, setup or anything else, and I will keep it simple.