Cairn: Pet Contract Lab

Inspiration

A custom Codex pet can look plausible while carrying a wrong schema version, unsafe identifier, incorrect atlas size, broken frame layout, undeclared files, mismatched hashes, or an unsafe installation target. Cairn makes those seams inspectable before installation.

What it does

Cairn: Pet Contract Lab is a local-first developer tool for Codex v2 pet packages. It validates the exact pet.json and spritesheet.webp package membrane, checks the 1536 × 2288 transparent atlas and 8 × 11 frame grid, and exposes all nine animation states and sixteen clockwise look directions in a static preview.

Its lifecycle is dry-run first. It sandbox-installs only into an explicitly supplied Codex home, verifies the exact payload and SHA-256 hashes, and can reversibly move the package into pets-shed without deleting it. A machine-readable receipt binds what the tool observed.

How we built it

Codex with GPT-5.6 helped develop and test the Python CLI, static HTML/JavaScript inspector, adverse-case fixtures, lifecycle guards, judge path, and documentation. Michael D. Travis selected the problem, product boundary, human gates, track, and every public release decision.

Challenges

The main challenge was making package state visible without pretending to observe live Codex telemetry. We also had to make filesystem consequences explicit, reversible, and testable while keeping all demonstrations away from a real Codex home.

Accomplishments and learning

The public-clone judge path passes from a credential-free checkout. Cairn rejects the deliberate invalid fixture, accepts the valid synthetic fixture, performs no network calls, and does not touch a real Codex home. We learned that inspectability and reversibility can turn an opaque asset-installation process into a reviewable developer contract.

What's next

Next steps include broader operating-system testing, richer fixture generation, visual-quality gates, and carefully governed support for original character packages after identity, motion, rights, and release review.

Built With

• codex
• css3
• github
• gpt-5.6
• html5
• javascript
• python