Ultracodex config

Skill 0xPiPo/ultracodex/skill/ultracodex-config

Offload Claude Code's heavy coding to Codex; keep judgment and verification on Claude.

Install

npx -y skills add 0xPiPo/ultracodex --skill ultracodex-config

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

View and change the persistent ultracodex worker defaults (model, reasoning effort, service tier, sandbox, timeout, web search, max parallel). Edits ~/.claude/skills/ultracodex/ultracodex.env, which codex-worker.sh sources for its defaults (per-call env vars still override). Invoke as /ultracodex-config, or when the user wants to configure ultracodex, change its default model/effort/speed tier, or set persistent worker settings.

SKILL.md

3.1 KB, 855 tokens by cl100k_base, as published. Nobody here has run it

/ultracodex-config

Manage the persistent ultracodex defaults. These live in ~/.claude/skills/ultracodex/ultracodex.env; codex-worker.sh sources that file, and any per-call ULTRACODEX_* env var still overrides a saved default.

Helper

bash ~/.claude/skills/ultracodex-config/ucx-config.sh show          # current settings
bash ~/.claude/skills/ultracodex-config/ucx-config.sh get   KEY
bash ~/.claude/skills/ultracodex-config/ucx-config.sh set   KEY VALUE
bash ~/.claude/skills/ultracodex-config/ucx-config.sh unset KEY     # revert to default

KEY may be given short (MODEL, EFFORT, TIER, SANDBOX, TIMEOUT, SEARCH, MAX, MCP) or full (ULTRACODEX_MODEL). Values are validated (a bad EFFORT/TIER/SANDBOX/MODEL/ MCP is rejected).

Settings

Key	Values	Default	Meaning
`MODEL`	`gpt-5.6-sol`, `gpt-5.6-terra`, `gpt-5.6-luna` (also legacy `gpt-5.5`/`gpt-5.4`/`gpt-5.4-mini`)	gpt-5.6-sol	Worker model. sol = flagship; terra = standard; luna = fast/cheap for mechanical/bulk.
`EFFORT`	`minimal`, `low`, `medium`, `high`, `xhigh`, `max`, `ultra`	xhigh	Reasoning effort (raises token volume, not rate). `ultra` spawns worker subagents — avoid (the skill fans out already).
`TIER`	`priority`, `default`	default	Service tier. priority = ~1.5x speed, more usage.
`SANDBOX`	`workspace-write`, `read-only`, `danger-full-access`	workspace-write	Worker sandbox.
`TIMEOUT`	integer seconds	600	Per-unit timeout.
`SEARCH`	`1` / unset	off	Allow live web search.
`MAX`	integer (0 = unlimited)	0	Max parallel workers in codex-batch.sh.
`MCP`	`context7`, `playwright` (comma-sep)	none	MCP servers for each worker. Prefer per-call — persisting loads it on every worker.

How to run it interactively

When invoked as /ultracodex-config:

Run ucx-config.sh show and report the current settings.
If the user named a change, apply it with set/unset and confirm.
If they didn't, ask (via AskUserQuestion) which setting to change and to what, then apply it. Common recipe: for a fast/cheap default, set MODEL=gpt-5.6-luna, EFFORT=low, TIER=priority; for a thorough default, MODEL=gpt-5.6-sol, EFFORT=xhigh.

Changes take effect on the next worker run. No restart needed (the file is sourced each invocation).

GSD model selector

The related GSD model selection lives separately in ~/.gsd/defaults.json (model_overrides): judgment agents on the flagship (e.g. gpt-5.6-sol), mechanical/ high-volume agents on the cheap model (e.g. gpt-5.6-luna), mirroring the ultracodex doctrine. Edit that file to retune which GSD agents get the cheap vs strong model.

What ships with it: 1 file

5.7 KB alongside SKILL.md, 1 of them executable

ucx-config.shruns5.7 KB