This skill should be used when the user asks to "create a codex prompt", "write a CTM", "format task for codex", "optimize prompt for codex", "submit to codex", or needs to generate a Codex Task Manifest (CTM) for delegation to OpenAI's Codex agent.
Generates Codex Task Manifests (CTM) for OpenAI's Codex agent when users request codex prompts or CTM formatting. It creates structured YAML with explicit constraints, verification requirements, and non-goals to prevent scope creep and ensure Codex delivers provable results.
/plugin marketplace add dkoosis/cc-plugins/plugin install codex@cc-pluginsThis skill inherits all available tools. When active, it can use any tool Claude has access to.
references/ctm-examples.mdreferences/ctm-template.ymlGuide for creating effective prompts for OpenAI's Codex coding agent using the Codex Task Manifest (CTM) format.
Codex is OpenAI's coding agent that operates in a sandboxed Linux environment. It excels at clearly bounded tasks but struggles without explicit guardrails. The CTM format provides structure that plays to Codex's strengths while preventing common failure modes.
The Codex Task Manifest counters Codex's weaknesses by:
| Section | Purpose |
|---|---|
task | Identity: issue, branch, deliverable |
role | Posture: conservative, craftsmanship |
context | Success definition, non-goals |
inputs | Relevant paths, test hints |
constraints | Design rules, Go rules |
plan | Required steps before editing |
verification | Proof of work requirements |
friction_report | Learning loop for future |
pr | PR template with checklist |
To generate a CTM from a GitHub issue:
codex/issue-[n]-[slug]To generate a CTM from a task nugget:
To generate a CTM from a plain description:
task:
id: "[gh-issue-number]"
title: "[short imperative summary]"
repo: "github.com/OWNER/REPO" # From .claude/codex.local.md
branch: "codex/issue-[n]-[slug]"
deliverable: "Open a PR that closes #[n]"
codex/issue-{number}-{slug}role:
primary: "Senior Go engineer working in this repo"
posture:
- "Conservative changes"
- "Craftsmanship over cleverness"
- "Assume long-term maintainership"
Codex mirrors the posture given. Explicit conservatism reduces overreach.
non_goals:
- "No architectural redesign"
- "No renames or formatting-only changes"
- "No new dependencies unless explicitly required"
Non-goals act as brakes. Without them, Codex often "improves" unrelated code.
plan:
required_steps:
- "Restate the issue requirements in your own words"
- "Identify current behavior and exact code locations"
- "Propose a minimal implementation approach"
- "Define test strategy"
- "List risks or ambiguities"
Codex is far more accurate after explicitly reasoning in writing.
verification:
must_provide:
- "List of commands run and their outcomes"
- "Summary of key code changes (what + why)"
- "Risk assessment and rollback notes"
- "Anything intentionally left out"
Forces Codex to prove work rather than declaring success early.
friction_report:
minimum_items: 3
categories:
- "Repo navigation"
- "Build/test workflow"
- "Code ergonomics / API design"
Turns PRs into a learning loop. Mine friction reports into Orca nugs (traps, rules) for future improvement.
For simple tasks, use a minimal CTM:
task:
title: "[imperative summary]"
repo: "github.com/OWNER/REPO" # From .claude/codex.local.md
branch: "codex/[slug]"
deliverable: "Open a PR"
context:
success_definition:
- "[specific criterion]"
non_goals:
- "No unrelated changes"
inputs:
relevant_paths_hint:
- "[path/to/relevant/code]"
testing:
commands:
- "mage qa"
- "go test ./..."
references/ctm-template.yml - Full CTM template with all sectionsreferences/ctm-examples.md - Real CTM examples for different task types| Task Type | Key CTM Focus |
|---|---|
| Bug fix | Clear repro steps, test strategy |
| QA/lint | File list, specific rules to enforce |
| Feature | Bounded scope, explicit non-goals |
| Review | Output format, criteria checklist |
Creating algorithmic art using p5.js with seeded randomness and interactive parameter exploration. Use this when users request creating art using code, generative art, algorithmic art, flow fields, or particle systems. Create original algorithmic art rather than copying existing artists' work to avoid copyright violations.
Applies 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.
Create beautiful visual art in .png and .pdf documents using design philosophy. You should use this skill when the user asks to create a poster, piece of art, design, or other static piece. Create original visual designs, never copying existing artists' work to avoid copyright violations.