ワンクリックで
nova-implement-plan
Implement strictly from an approved plan. Requires PLAN_INPUT_PATH and PLAN_APPROVED=true before execution.
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。
メニュー
Implement strictly from an approved plan. Requires PLAN_INPUT_PATH and PLAN_APPROVED=true before execution.
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 | nova-implement-plan |
| description | Implement strictly from an approved plan. Requires PLAN_INPUT_PATH and PLAN_APPROVED=true before execution. |
| license | MIT |
| allowed-tools | Read Glob Grep Write Edit |
| disallowed-tools | NotebookEdit |
| user-invocable | true |
| disable-model-invocation | true |
| compatibility | Designed for Claude Code; project validation through Bash follows the normal permission flow. |
| metadata | {"nova-user-invocable":"true","nova-model-invocable":"false","nova-subagent-safe":"true","nova-destructive-actions":"medium"} |
| argument-hint | Example: implement-plan PLAN_INPUT_PATH=docs/plans/refund.md PLAN_APPROVED=true |
This file is the supporting behavioral contract for /nova-plugin:implement-plan and the deprecated /nova-plugin:nova-implement-plan compatibility entrypoint. Prefer the direct command; the compatibility name remains only for the current major-version migration window.
KEY=value inputs using ../_shared/parameter-resolution.md; explicit non-conflicting values take precedence.../_shared/safety-preflight.md before side effects. Never infer approval, destructive scope, credentials, or output destinations.../_shared/output-contracts.md and ../_shared/artifact-policy.md; report completed, skipped, and blocked validation truthfully.$ARGUMENTS against the workflow-specific inputs below.Generated from
workflow-specs/behaviors.json. This block is authoritative. Runnode scripts/generate-behavior-surfaces.mjs --writeafter changing the IR; if explanatory text below conflicts, fail closed.
PLAN_INPUT_PATH(required aliases=PLAN_PATH); PLAN_APPROVED(required aliases=APPROVED exact=true,"true")validate-approval → load-plan → execute → verify → traceworkspace-and-chat; order=plan-step trace → validation → deviations; severity=none.approval-required; failure order=status → completed plan steps → blocked plan step → blocker → safe next action.runtime/contracts/implement-plan.json#behaviorContract embeds the complete decision table, invariants, stops, field definitions, validation, and failure contract from the same source. Detailed guidance below may not override it.Execute approved plan steps with minimal deviation and clear traceability.
| Parameter | Required | Default | Notes | Example |
|---|---|---|---|---|
PLAN_INPUT_PATH | Yes | N/A | Approved plan file | docs/plans/refund.md |
PLAN_APPROVED | Yes | N/A | Must be exactly true | true |
Use implement-plan with the approved plan in docs/plans/login.md.implement-plan PLAN_INPUT_PATH=docs/plans/login.md PLAN_APPROVED=true.You are Claude Code acting as a disciplined software engineer.
This command performs implementation ONLY, strictly based on an approved plan document.
No exploration. No redesign. No scope expansion.
IMPLEMENT CODE STRICTLY BASED ON AN APPROVED PLAN
From $ARGUMENTS, extract:
PLAN_INPUT_PATH: <PLAN_INPUT_PATH>
If PLAN_INPUT_PATH is missing:
PLAN_APPROVED: <PLAN_APPROVED>
Rules:
true (case-sensitive)If PLAN_APPROVED is missing or not exactly true:
PLAN_APPROVED=true is required to proceedYou MUST:
PLAN_INPUT_PATH before any implementationYou MUST NOT:
Deviation is allowed ONLY if:
If a deviation occurs, you MUST:
If deviation is non-trivial:
While implementing, you MUST:
You MUST NOT:
Implementation should maintain clear traceability:
Provide the following in chat output:
Implemented code changes
A short implementation summary
Explicit deviation notes (if any)
Do NOT:
Generate a Java/Spring backend design plan with mandatory 12 sections and write to PLAN_OUTPUT_PATH.
Run a semi-automated Codex review -> Claude Code fix -> local checks -> Codex verify closure loop for the current branch. Use when Claude Code should orchestrate review-driven fixes with external Bash scripts, structured review artifacts, validation reports, merge readiness, and residual-risk summaries.
Run Codex-only branch review with external scripts and structured review artifacts. Use when Claude Code should collect current branch diff context and ask Codex to produce a high-confidence review report without modifying code.
Run Codex verification against an existing review.md and the current branch state. Use when Claude Code or a human has already applied fixes and only needs a focused verify pass, merge recommendation, and residual blockers.
Lightweight observer-style exploration for quick understanding alignment.
Reviewer-style exploration focused on questions and risk signals, without proposing fixes.