smixs/disruptor-skills — explained in plain English
Analysis updated 2026-05-18
Install the skill set so an AI coding agent follows a structured design, spec, and review pipeline instead of jumping straight to code.
Use the router skill to send a bug report straight to the QA stage rather than starting a fresh design.
Run the fleet lane skill to have parallel subagents build a fast MVP from a rough idea.
Use the unvibe review skill to push back on a codebase that has grown bloated.
| smixs/disruptor-skills | aim-uofa/reasonmatch | airbone42/360-data-athlete | |
|---|---|---|---|
| Stars | 12 | 12 | 12 |
| Language | Python | Python | Python |
| Setup difficulty | easy | hard | hard |
| Complexity | 2/5 | 5/5 | 4/5 |
| Audience | developer | researcher | general |
Figures from each repo's GitHub metadata at analysis time.
Installs as a set of Markdown skill files, no build step, though it assumes a skills capable agent like Claude Code, Codex, or Cursor.
Disruptor is a collection of twelve linked skills meant to guide an AI coding agent, such as Claude Code, Codex, or Cursor, through a full software delivery process instead of letting it freestyle from a vague prompt. The README frames its core idea as distrust of an agent's own sense of being done, arguing that unread tests, specs with hidden defaults, and rubber stamped reviews are common failure points that this pipeline is designed to catch. The skills are organized as stages, starting with a one time setup that records repo conventions and a shared glossary, followed by a structured design interview, a formal specification step, executable architecture rules enforced by a linter and import graph checks, breaking work into small shippable vertical slices, and a build and review loop that keeps going until findings settle down. That loop includes independent reviewer skills that look at the work with fresh context, and a separate skill whose job is to cut and simplify rather than add more. Later stages cover demo quality testing and a guided server setup and deploy step. A router skill decides which stage a given task should enter based on how the request is phrased, such as describing a new idea, reporting a bug, or saying a codebase feels bloated. Some skills trigger automatically when the agent recognizes a matching situation, while others are invoked directly as slash commands. Installation options include a dedicated skills installer, a shell script that copies the skill folders into a local skills directory, or a plain git clone. The project can also be installed as a Claude Code plugin. According to the README, a user does not need to memorize the stages, since describing the task in plain language, such as asking to design something properly or to review until it converges, is enough for the router to place the work at the right stage. The project is released under the MIT license.
A set of 12 linked Claude Code style skills that route an idea through design, spec, review, and deploy stages instead of letting an agent freestyle.
Mainly Python. The stack also includes Markdown, YAML.
MIT licensed, so the skills can be used, copied, and modified freely, including commercially, as long as the copyright notice is kept.
Setup difficulty is rated easy, with roughly 5min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Don't trust strangers blindly. Verify against the repo.