ultracook

Orchestrate a seven-phase cheese pipeline with fresh-context sub-agent isolation.

15|1|Updated May 4, 2026
One-click install
npx skills add https://github.com/paulnsorensen/easy-cheese --skill ultracook
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: ultracook
Source: https://github.com/paulnsorensen/easy-cheese/tree/main/skills/ultracook
Command: npx skills add https://github.com/paulnsorensen/easy-cheese --skill ultracook

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

It eliminates the need for step-by-step approval when transforming a locked spec through a multi-phase cheese pipeline, while preventing cross-phase context contamination.

Core Features & Use Cases

  • Fresh-context phase isolation: Runs cook → press → age → cure → age → cure → age in separate sub-agents so each phase reasons without relying on previous phase internal context.
  • Deterministic sequential orchestration: Executes the full seven-phase chain with --auto semantics propagated end-to-end and stops on explicit halt conditions.
  • Durable handoff artifacts: Writes per-phase handoff reports to .cheese/<phase>/<slug>.md and uses them to drive continuation and early stopping.
  • Controlled cure passes: Enforces the two-cure-pass limit via the fixed chain length rather than relying on age counting.
  • Terminal PR workflow: Pushes to an already-open PR at the end, or opens a new PR only when --open-pr is specified.

Quick Start

Run ultracook on your spec slug with fresh-context isolation and automatic pipeline execution by typing: /ultracook your-slug.

Frequently Asked Questions about ultracook

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I automate a multi-phase workflow pipeline without approving each step?

You can automate a multi-phase workflow pipeline end-to-end using fresh-context sub-agent isolation with `--auto` execution semantics. This approach eliminates per-step approvals by orchestrating the entire chain sequentially and stopping only when explicit halt conditions are met.

How does fresh-context isolation prevent context contamination in iterative refinement?

Fresh-context isolation prevents cross-phase context contamination by spawning full-peer sub-agents for each pipeline phase. Each sub-agent reasons independently without relying on previous phase internal context, ensuring durable phase handoffs drive continuation and early stopping.

Can I enforce a specific number of iterative refinement passes in a spec-driven automation pipeline?

Yes, you can enforce a specific number of iterative refinement passes by using a fixed chain length. This approach caps cure passes by structural design rather than relying on dynamic age counting, ensuring deterministic sequential orchestration.

Do I need an existing GitHub PR to run an automated workflow pipeline to completion?

You do not need an existing GitHub PR; the pipeline can push to an already-open PR at the end. If no PR exists, it will only open a new PR when the optional `--open-pr` flag is explicitly specified.

What are the limitations of using fixed chain length for workflow pipeline halt-aware stopping?

The fixed chain length limits workflow pipeline halt-aware stopping by strictly enforcing a two-cure-pass cap through structural design. You cannot dynamically alter the pipeline length or add extra iterative refinement phases beyond the predefined seven-phase chain.