← Back to Skills Marketplace
hanjo92

Codex Goal Decomposer

by Song Seung Hu · GitHub ↗ · v0.1.0 · MIT-0
cross-platform ✓ Security Clean
48
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install codex-goal-decomposer
Description
Break a large, fuzzy, or long-running objective into smaller goal-mode tasks with clear scope, stop conditions, and validation steps. Use when the user wants...
README (SKILL.md)

Goal Decomposer

Turn one big goal into a short sequence of smaller goals that are concrete enough for goal mode to handle well.

Core rule

Do not send one vague multi-day objective into goal mode if it can be split into clearer checkpoints.

Workflow

  1. Read the user's real objective.
  2. Restate the final desired outcome in one sentence.
  3. Split it into 3-7 smaller goals.
    • Prefer goals that are roughly 20-120 minutes of focused work.
    • Keep each goal to one main deliverable or checkpoint.
    • Preserve dependencies and execution order.
  4. For each smaller goal, define:
    • title
    • why it exists
    • scope / constraints
    • done condition
    • validation method
    • the exact /goal ... command to start it
  5. If the original request is still too vague, ask for the one missing decision that blocks useful decomposition.
  6. If the task is too small for goal mode, say so and recommend normal execution instead.

Heuristics

A good sub-goal should have:

  • one clear outcome
  • a visible checkpoint
  • a concrete validation step
  • an obvious stop condition

Avoid sub-goals like:

  • "keep improving the app"
  • "clean up the codebase"
  • "make it production ready"

Prefer sub-goals like:

  • "make the project build cleanly on Node 22"
  • "replace the old auth helper and pass auth tests"
  • "write README setup steps and verify them on a fresh clone"

Output format

Use this structure unless the user requests another format.

Final objective

One sentence.

Recommended sub-goals

For each goal, include:

  1. Goal title
    • Why: ...
    • Scope: ...
    • Done when: ...
    • Validate with: ...
    • Command: /goal ...

Suggested execution order

Use a numbered list.

Notes

  • Mention risky or ambiguous goals.
  • Call out any goal that should be split again before execution.
  • Say when a normal one-turn task is better than goal mode.

Goal command style

Write /goal commands so they include:

  • the objective
  • what to read first
  • what not to change when relevant
  • how to validate
  • the stopping condition

Keep each command compact but specific.

Good pattern:

/goal Update the README setup section for this repo. Read README.md and package.json first. Do not change application code. Validate by checking the documented commands against the repo files. Stop when the README has accurate install, run, and example usage steps.

Boundaries

  • Do not claim goal mode can handle arbitrary open-ended work well.
  • Do not produce a giant backlog; produce the next useful chain.
  • If the user wants execution too, first decompose, then recommend starting with the first sub-goal.
Usage Guidance
This appears safe to install as a planning aid. Review any generated /goal command before running it, because the later tasks may ask Codex to read or modify project files depending on your original objective.
Capability Assessment
Purpose & Capability
The stated purpose is goal decomposition, and the artifact only provides instructions for producing scoped sub-goals, validation steps, and /goal command text.
Instruction Scope
Instructions are bounded to planning and formatting; they emphasize scope, stop conditions, validation, and asking a clarifying question when needed.
Install Mechanism
The package contains only SKILL.md, no executable scripts, declared dependencies, install hooks, or bundled tooling.
Credentials
No artifact evidence requests network access, local credential access, broad indexing, background work, or mutation authority beyond user-directed future /goal tasks.
Persistence & Privilege
No persistence mechanism, privilege escalation, background agent, cron job, credential storage, or session/profile use is present.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install codex-goal-decomposer
  3. After installation, invoke the skill by name or use /codex-goal-decomposer
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v0.1.0
Initial release: decompose large objectives into smaller Codex /goal tasks with done conditions and validation steps.
Metadata
Slug codex-goal-decomposer
Version 0.1.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is Codex Goal Decomposer?

Break a large, fuzzy, or long-running objective into smaller goal-mode tasks with clear scope, stop conditions, and validation steps. Use when the user wants... It is an AI Agent Skill for Claude Code / OpenClaw, with 48 downloads so far.

How do I install Codex Goal Decomposer?

Run "/install codex-goal-decomposer" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Codex Goal Decomposer free?

Yes, Codex Goal Decomposer is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does Codex Goal Decomposer support?

Codex Goal Decomposer is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created Codex Goal Decomposer?

It is built and maintained by Song Seung Hu (@hanjo92); the current version is v0.1.0.

💬 Comments