Skip to content

btseytlin/ultrapack / skills/make

Orchestrate the full Ultrapack workflow in Codex and Pi: create or resume a task file, design, plan, execute, verify, review, and validate the observable goal. Use when the user asks to build, fix, or deliver a non-trivial change with the complete workflow.

FileKindSize
SKILL.mdskill md3.4 KB
---
name: make
description: "Orchestrate the full Ultrapack workflow in Codex and Pi: create or resume a task file, design, plan, execute, verify, review, and validate the observable goal. Use when the user asks to build, fix, or deliver a non-trivial change with the complete workflow."
---

# Ultrapack workflow

Use this as the Codex and Pi equivalent of Claude Code's `/up:make`. Treat the user's remaining message as the task description. If the first token is exactly `handsoff`, enable hands-off mode and remove that token before making the slug. Otherwise keep the task description unchanged.

<required>
Before responding or writing the task file, read the [brevity rules](../_brevity.md) and [global principles](../_principles.md). Apply them throughout the workflow.
</required>

1. Derive a kebab-case slug of at most three words. If `docs/tasks/<slug>.md` exists, read its status and resume at the next incomplete stage. If several tasks are in flight, ask which one to resume.
2. For a new task, create `docs/tasks/<slug>.md` from this template. Set Status to `design`, Branch to `main`, Worktree to `none`, Mode from the argument, and make Goal a draft observable outcome from the request.

```markdown
# <Task Title>

**Status:** design
**Branch:** main
**Worktree:** none
**Goal:** <observable end state; who uses the result and for what; Useless if: cheap results that pass the checks and defeat that purpose, per GPC9>
**Mode:** <interactive|hands-off>

## Context
## Design
### Invariants
### Principles
### Assumptions
### Unknowns
## Plan
## Verify
## Code smells
## Conclusion
### Hands-off decisions
### Deferred (needs user input)
```
3. Classify the task. Skip Design for trivial changes without confirmation. In interactive mode, confirm before skipping Plan or skipping Design for a non-trivial change. In hands-off mode, default to Medium unless the task is unmistakably a one-line change.
4. Run `udesign` unless skipped, then set Status to `planning`. Follow the existing branch and worktree rules without another approval. In hands-off mode, use `git-worktrees` and stop rather than edit the main branch if it cannot provision isolation.
5. Run `uplan` unless skipped, then set Status to `executing`. In hands-off mode, record that the plan was auto-approved.
6. Run `uexecute`, then `uverify`. Send demonstrated failures back to execution until the verification attacks pass. Set Status to `reviewing` and run `ureview`.
7. Set Status to `validating`. Mark `done` only when the task file's Goal has direct end-to-end evidence or the user confirms an outcome that the harness cannot observe. Do not equate a clean diff, passing tests, review, or a met number with a confirmed Goal. Check how the result met it against the Goal's `Useless if:` list, per GPC9.
8. Once done, scan `CLAUDE.md`, `README.md`, and non-task project documentation for material updates. Present branch merge/PR, worktree cleanup, or defer options; never push or merge without the user's choice.

Use `handsoff` for the shared safety contract. Record every automatic decision in `### Hands-off decisions`; never invent a value without a conservative, explicit basis. Ask the user only when there is no safe default or a blocker prevents the next stage.

## Harness adaptation

Use the existing skills by name (`udesign`, `uplan`, `uexecute`, `uverify`, `ureview`, and `git-worktrees`). Codex users invoke `$make`. Pi users invoke `/make` or `/skill:make`.
brainstormingobra/superpowersYou MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or planning anything new, in software or out of it (a talk, a business, a renovation).Stars: 322kUpdated Claude CodeCodexVerifiedInstall
brand-guidelinesanthropics/skillsApplies Anthropic's official brand colors and typography to any sort of artifact that may benefit from having Anthropic's look-and-feel. Use it when brand colors or style guidelines, visual formatting, or company design standards apply.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
frontend-designanthropics/skillsGuidance for distinctive, intentional visual design when building new UI or reshaping an existing one. Helps with aesthetic direction, typography, and making choices that don't read as templated defaults.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
mcp-builderanthropics/skillsGuide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK).Stars: 198kUpdated Claude CodeCodexVerifiedInstall