メニュー
Use when executing a written implementation plan - per step read, change, verify, mark done; loop back on failure
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
| name | executing-a-plan-step-by-step |
| description | Use when executing a written implementation plan - per step read, change, verify, mark done; loop back on failure |
| kind | workflow |
| audience | dev |
| ai-tools | any |
| complexity | chained |
| time | depends on plan |
| version | 1.0.0 |
| source | bundled |
| composes | ["writing-an-implementation-plan"] |
Executes an implementation plan step-by-step with discipline: read what the step says, make the change, run the step's verification, mark done. On verification failure, stop and diagnose before moving on.
A completed implementation, plus an updated plan document with each step marked done and any deviations recorded.
This is a chained workflow. Each step is a sub-cycle.
Open the plan. Find the first step that's not marked done. If the plan has no marks yet, start at step 1.
If you're resuming after a break, re-read the steps that were already done — you need that context loaded back into your head.
For each step:
What does this step say to do? What does it say is the verification? What files does it touch?
If the step is unclear, do NOT improvise. Update the plan first to clarify, then execute the clarified step.
Before touching anything, read the file(s) the step affects in their current state. Skipping this is the #1 cause of "AI confidently wrote the wrong thing."
Make the smallest change that satisfies the step. Do NOT make adjacent improvements ("while I'm here..."). Adjacent changes go in a side notebook for later, not in this commit.
Run exactly the verification the plan specified. Capture the output.
If the step left the system in a working state, commit. Use a message that references the step.
When verification fails:
If the failure reveals the plan was wrong, update the plan before continuing. Don't silently work around the plan.
Maintain a "deferred" list as you go: things you noticed that should change later, but aren't part of this plan. This is where "while I'm here..." goes. You'll address it after the plan is done (or never, which is also fine).
When all steps are marked done:
Input: A 12-step implementation plan for SSO.
Process:
Use when the meta-builder routes to kind=discipline. Generates a discipline spell with the rule, excuses table, warning signs, and hard gates.
Use when the user wants to capture a repeatable task as a reusable spell, or asks "how do I build a workflow / skill / framework for X"
Use when the meta-builder routes to kind=subagent. Generates a spell that dispatches one or more subagents to do work in parallel or in isolation.
Use when the meta-builder routes to kind=workflow. Generates a workflow spell with explicit stages, checkpoints, and loop-back conditions.
Use when composing multiple spells into a chain that runs end-to-end (e.g. brainstorm -> plan -> execute -> verify).
Use when browsing what spells are available - bundled seeds, your personal library, or both. Filter by kind, audience, or update status.
