clairvoyance

Clairvoyance enhances AI agents' software design capabilities with strategic insights and diagnostic skills.

<p align="center"> <img src="docs/src/assets/crystal-ball.svg" width="80" alt="Clairvoyance"> </p> <h1 align="center">Clairvoyance</h1>

AI agents can write working code, but they don't stop to consider effective design unless asked. Clairvoyance is a set of skills inspired by John Ousterhout's A Philosophy of Software Design. Each skill gives your agent extrasensory perspective around software design, with concrete tests to see ahead of obstacles during implementation and review.

How It Works

Skills activate automatically and push your agent to ask questions like:

Does this interface hide real complexity, or just pass things through?
Can this method be understood without reading another one in a different file?
Is error handling pushing work onto callers that the module could handle itself?

You can also invoke them directly. Use /red-flags to trigger a design smell scan, /deep-modules to check interface depth and /design-it-twice to compare alternatives before committing.

Installation

Note: Installation differs by platform. Claude Code has a built-in plugin marketplace. Codex and OpenCode require manual setup.

Claude Code (via Plugin Marketplace)

/plugin marketplace add codybrom/clairvoyance
/plugin install clairvoyance@clairvoyance-plugins

skills.sh

npx skills add codybrom/clairvoyance --skill '*'

Codex

Tell Codex:

Fetch and follow instructions from https://raw.githubusercontent.com/codybrom/clairvoyance/refs/heads/main/.codex/INSTALL.md

See .codex/INSTALL.md for detailed steps.

OpenCode

Tell OpenCode:

Fetch and follow instructions from https://raw.githubusercontent.com/codybrom/clairvoyance/refs/heads/main/.opencode/INSTALL.md

See .opencode/INSTALL.md for detailed steps.

llms.txt

Machine-readable skill index for LLM agents:

clairvoyance.fyi/llms.txt — table of contents with descriptions
clairvoyance.fyi/llms-full.txt — full content of all skills

What's Inside

Structure & Modules

Skill	Covers
`deep-modules`	Module depth, shallow modules, classitis, pass-through methods, interface vs implementation
`module-boundaries`	Merge vs split, conjoined methods, method splitting, dependency minimization
`information-hiding`	Information leakage, temporal decomposition, partial hiding, false encapsulation
`pull-complexity-down`	Caller burden, configuration parameters, the core asymmetry

Abstraction & Generality

Skill	Covers
`abstraction-quality`	Genuine vs false abstractions, layer boundaries, decorators
`general-vs-special`	Interface generality, special-general mixture, edge-case elimination
`error-design`	Define errors out of existence, exception masking, aggregation, just crash

Clarity & Communication

Skill	Covers
`naming-obviousness`	Isolation test, scope-length principle, consistency, avoid extra words
`comments-docs`	Comment types, comments-first workflow, cross-module documentation

Process & Evolution

Skill	Covers
`strategic-mindset`	Strategic vs tactical, investment rule, tactical tornado
`design-it-twice`	Generate alternatives, compare on criteria, synthesize
`code-evolution`	"Designed this way" standard, repetition, technical debt
`complexity-recognition`	Change amplification, cognitive load, unknown unknowns

Diagnostic

Skill	Covers
`red-flags`	Design smell scan covering structure, boundaries, documentation, naming, and process
`design-review`	Structured review funnel from complexity triage through structural, interface, and surface checks
`diagnose`	Routes a vague symptom or complaint to the most relevant skill via a decision tree

Attribution

These skills are adapted in part from the teachings of John Ousterhout, professor of computer science at Stanford University, and his book A Philosophy of Software Design. This project is not affiliated with, endorsed by, or sponsored by John Ousterhout, Stanford University, or the publishers of A Philosophy of Software Design.

If you find these skills useful, you should really buy and read the book. The skills in this repo are by no means a substitute for reading it. It is the definitive treatment of these ideas and an enjoyable read for any dev. Available from Amazon (no affiliate link). Also available in German (O'Reilly, 2021) and Chinese (Posts and Telecommunications Press, 2024).

The skills and code in this project are independently authored original works by the project's contributors. Brief quotations from the book are sometimes used with full attribution for purposes of commentary, criticism, and education. All trademarks and copyrights are the property of their respective owners.

Contributing

Contributions are welcome beyond the inspiration material, but should reinforce the core philosophy of thinking strategically about software design.

To contribute:

Fork the repository
Create a branch
Follow the writing-skills skill from Superpowers for creating and testing skills
Add your skill in skills/<skill-name>/SKILL.md with optional references/ files
Commit your changes and submit a PR

Updating

Claude Code: /plugin update clairvoyance
skills.sh: npx skills update codybrom/clairvoyance
Codex / OpenCode: Re-run the install instructions — these agents fetch the latest from main each time.
llms.txt: Always up to date at clairvoyance.fyi/llms-full.txt.