ethanplusai/codex-on-crack/crack-setup

Claude Code and agent skill by ethanplusai/codex-on-crack

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Configure or change the approved worker models for codex-on-crack, inspect available models and compatibility, preview role files, or undo a setup. The user selects the lead in Codex. Setup preserves root settings, credentials, unrelated agents, and config.toml.

Install

npx vskill@latest install ethanplusai/codex-on-crack/crack-setup

SKILL.md

Choose workers; preserve the lead

Helpers live in ../crack/scripts/. Keep the user's existing model/provider choices. Installing skills alone does not configure workers or change the lead.

  1. Run node ../crack/scripts/setup.mjs scan with the session's home/profile.

Show relevant eligible models, efforts, image support, context, and exclusions. Catalog presence is not account access or a working provider route. Do not rank capabilities the catalog does not establish or invent prices.

  1. Read an existing crack.toml before proposing changes. Start with one builder

chosen by the user. Additional roles can map to implementation, ui, review, research, or tests with a tasks array. These are user preferences, not demonstrated model strengths. Set the exact model and a supported effort; keep fallback optional. Never change the selected lead to make setup pass.

  1. Write a draft in a task-owned directory. Preview with

node ../crack/scripts/setup.mjs plan --roles DRAFT and show the concrete model/effort and file changes. Apply the agreed configuration with node ../crack/scripts/setup.mjs apply --roles DRAFT. If the specific model choices or global writes aren't yet authorized, obtain that decision after producing the preview. Don't repeat approval already given for those choices.

  1. Return the receipt path and explain that a fresh client session must load the

generated native roles. Run doctor.mjs; then use a real authorized task to verify child metadata. No paid installation smoke test is needed.

Draft shape (replace the model with an exact eligible ID):

toml
schema_version = 1
[roles.builder]
model = "MODEL_FROM_SCAN"
writes = true
tasks = ["implementation", "tests"]
brief = "Implements the scoped task and verifies it for lead review."

Role names use lowercase letters/digits/underscores and cannot end _fallback. Read-only roles set writes = false. Use capability/route checks, not a model's brand, to determine whether a proposed setup is compatible. A model recommendation does not grant permission to share private source with a new provider.

The current adapter conservatively checks the inherited parent provider. Native-model and Router-model catalog entries may both exist without a compatible mixed-provider session. Stop and explain an incompatible route; do not rewrite endpoints, credentials, or permissions. Custom-role model settings take precedence over a spawn override, so verify the role actually names the selected model.

Default to no global policy during initial setup. If an old orchestrator policy is active, explain the conflict and use an authorized --migrate-legacy preview rather than leaving both workflows active. This backs up/removes only its marked policy block; old role and skill files remain. Do not migrate merely to inspect or compare the packages.

Use --profile consistently; setup records profile/catalog for future checks. An offline --model-catalog can supply metadata only when it does not conflict with the configured catalog. Setup never refreshes a provider or reads credentials. Undo previews with setup.mjs undo --receipt PATH; --apply restores only when no later edits would be overwritten. Preserve receipts and hand-edited files.

Security

  • Security scanPassed, 90 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

7-day trend score
5

Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/ethanplusai/codex-on-crack/crack-setup/badge)](https://verified-skill.com/skills/ethanplusai/codex-on-crack/crack-setup)

Something wrong with this skill or its page? Tell us.