agentsclimarketplace

Planning first

Skill radustefandumitru/overdrive/skills/planning-first

Improve your AI coding agents easily.

Install
npx -y skills add radustefandumitru/overdrive --skill planning-first

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

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 13 stars13 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

Use for complex coding work on Codex, Cursor, Gemini, Antigravity, or other agents without Claude Code's native opusplan flow. Trigger on multi-file implementation, refactors, new features beyond one function, vague specs, migrations, release work, or tasks needing Explore -> Plan -> Implement -> Verify discipline. Pair after clarify-and-plan when ambiguity exists. Avoid trivial fixes, factual answers, and one-line edits.

SKILL.md

2.7 KB, 493 tokens by cl100k_base, as published. Nobody here has run it

Planning First

Use this skill when the implementation is large enough that rushing straight into edits is likely to create churn. It creates a portable version of a planning-first workflow for agents that do not have Claude Code's native /model opusplan or /ultraplan commands.

Runtime Choice

  • Claude Code: prefer native /model opusplan for multi-step coding and /ultraplan for complex multi-system tasks when available.
  • Codex, Cursor, Gemini CLI, Antigravity, shared .agents, or local project agents: use this skill.
  • If the user explicitly asks to skip planning, keep only a one-paragraph plan and proceed.

Workflow

  1. Explore:
    • Inspect the repo, instructions, relevant files, tests, scripts, and existing patterns.
    • Identify unknowns, constraints, and blast radius.
    • Use clarify-and-plan first if ambiguity would change the work.
  2. Plan:
    • Produce a short ordered plan with verification after each phase.
    • Name files or modules likely to change when known.
    • State what is intentionally out of scope.
  3. Implement:
    • Make the smallest complete changes that satisfy the plan.
    • Keep diffs surgical and preserve user-owned changes.
    • Update the plan as phases complete.
  4. Verify:
    • Run the narrowest meaningful tests/checks first.
    • Add browser, build, lint, typecheck, or manual verification when the risk justifies it.
    • If verification fails, loop on the failure before declaring completion.
  5. Close:
    • Summarize what changed, what passed, and any residual risk.

Planning Rules

  • Plans should usually be 3-6 steps, not a dissertation.
  • One in-progress step at a time.
  • If implementation reveals a better route, update the plan and explain the pivot.
  • For destructive, publishing, auth, billing, production-data, or account-affecting actions, ask before acting.

Hard Avoids

  • Do not use for one-line fixes, short explanations, or quick factual answers.
  • Do not create architecture documents unless the task needs them.
  • Do not keep planning after the user has already approved a clear plan and expects implementation.

Attribution

Original Overdrive skill by Stefan / Radu Stefan Dumitru, designed to give non-Claude agents a lightweight planning-first workflow.

What ships with it: 1 file

229 B alongside SKILL.md

agents/

Keep looking

Skills are one crate of 327,069. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.