Follow the phases below to implement the Plan, then Report the completed work.
IMPORTANT: If no plan is provided (i.e., $ARGUMENTS is empty or blank), STOP immediately and ask the user to provide one. Example usage: /implement-plan specs/new-feature.md
Plan
$ARGUMENTS
Phase 1: Understand Before Acting
Before writing any code:
Read the ENTIRE plan from start to finish
Determine scope from the plan's Scope section (API-only / Web-only / Full-stack / Shared)
Load relevant context based on scope — this is critical to implementing patterns correctly:
API scope: read <root>/.claude/context/api/00-overview.md + any specific pattern files the plan references
Web scope: read <root>/apps/web/CLAUDE.md
Full-stack: read both
Shared: read <root>/packages/shared/CLAUDE.md
Detect database work: if the plan mentions migrations, new tables, schema changes, or tableTypes.d.ts — flag this now. You'll need to invoke /create-database-migration at the appropriate step (never create migration files manually).
Build a todo inventory using TodoWrite, based on the "Step by Step Tasks" section (preferred) or "Files to Deliver" as fallback:
One todo per step from the plan (not per file — preserve the plan's ordering intent)
Add a final todo: "Final verification"
Phase 2: Implement Step by Step
Work through the plan's Step by Step Tasks in order. For each step:
Mark the step todo as in_progress
If the step introduces new functionality:
Write the test(s) first — they should fail initially (no production code yet)
Implement the production code to make the tests pass
Run the step's specific test command from the plan
If the step is structural (adding types, registering handlers, updating config with no new logic):
Implement directly, no test file needed
Run the incremental quality check for the affected app after each step:
Web: pnpm --filter web typecheck && pnpm --filter web lint
API: pnpm --filter api typecheck && pnpm --filter api lint
Shared: pnpm --filter shared build && pnpm --filter api typecheck && pnpm --filter web typecheck
Fix any type or lint errors before moving to the next step
Mark the step todo as completed
The reason to work step by step (rather than writing all tests then all production code) is that the plan's ordering encodes real dependencies — implementing out of order causes cascading failures that are harder to debug.
Phase 3: Final Verification
Re-read the plan's "Files to Deliver" section
Cross-check: does every listed file exist?
Cross-check: does every listed test case exist and pass?
Run ALL validation commands from the plan's "Validation Commands" section
If anything is missing, go back and complete it
Invoke the code-reviewer skill
Report
Only after Phase 3 is complete:
Summarize the work in a concise bullet point list
Show files and lines changed: git diff --stat
Confirm: "All tests pass" and "All validation commands pass"
Converted and distributed by TomeVault — claim your Tome and manage your conversions.
1---2name: incubateur-ademe-benefriches-implement-plan3description: Implement Plan4---56# Implement Plan78Follow the phases below to implement the `Plan`, then `Report` the completed work.910IMPORTANT: If no plan is provided (i.e., $ARGUMENTS is empty or blank), STOP immediately and ask the user to provide one. Example usage: `/implement-plan specs/new-feature.md`1112## Plan13$ARGUMENTS1415---1617## Phase 1: Understand Before Acting1819Before writing any code:20211. **Read the ENTIRE plan** from start to finish222. **Determine scope** from the plan's Scope section (API-only / Web-only / Full-stack / Shared)233. **Load relevant context** based on scope — this is critical to implementing patterns correctly:24 - API scope: read `<root>/.claude/context/api/00-overview.md` + any specific pattern files the plan references25 - Web scope: read `<root>/apps/web/CLAUDE.md`26 - Full-stack: read both27 - Shared: read `<root>/packages/shared/CLAUDE.md`284. **Detect database work**: if the plan mentions migrations, new tables, schema changes, or `tableTypes.d.ts` — flag this now. You'll need to invoke `/create-database-migration` at the appropriate step (never create migration files manually).295. **Build a todo inventory** using TodoWrite, based on the "Step by Step Tasks" section (preferred) or "Files to Deliver" as fallback:30 - One todo per step from the plan (not per file — preserve the plan's ordering intent)31 - Add a final todo: "Final verification"3233---3435## Phase 2: Implement Step by Step3637Work through the plan's **Step by Step Tasks** in order. For each step:38391. Mark the step todo as `in_progress`402. **If the step introduces new functionality:**41 - Write the test(s) first — they should fail initially (no production code yet)42 - Implement the production code to make the tests pass43 - Run the step's specific test command from the plan443. **If the step is structural** (adding types, registering handlers, updating config with no new logic):45 - Implement directly, no test file needed464. Run the incremental quality check for the affected app after each step:47 - Web: `pnpm --filter web typecheck && pnpm --filter web lint`48 - API: `pnpm --filter api typecheck && pnpm --filter api lint`49 - Shared: `pnpm --filter shared build && pnpm --filter api typecheck && pnpm --filter web typecheck`505. Fix any type or lint errors before moving to the next step516. Mark the step todo as `completed`5253The reason to work step by step (rather than writing all tests then all production code) is that the plan's ordering encodes real dependencies — implementing out of order causes cascading failures that are harder to debug.5455---5657## Phase 3: Final Verification58591. Re-read the plan's "Files to Deliver" section602. Cross-check: does every listed file exist?613. Cross-check: does every listed test case exist and pass?624. Run ALL validation commands from the plan's "Validation Commands" section635. If anything is missing, go back and complete it646. Invoke the `code-reviewer` skill6566---6768## Report6970Only after Phase 3 is complete:71721. Summarize the work in a concise bullet point list732. Show files and lines changed: `git diff --stat`743. Confirm: "All tests pass" and "All validation commands pass"7576---77> Converted and distributed by [TomeVault](https://tomevault.io/claim/incubateur-ademe) — claim your Tome and manage your conversions.78<!-- tomevault:4.0:skill_md:2026-04-11 -->
Run npx skillmds@latest add tomevault-io/incubateur-ademe-benefriches-implement-plan in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.
Implement Plan It is listed under Coding & Dev Tools on SkillMD.
This skill has not completed SkillMD's automated safety review yet. Independent scanners report: SkillSpector: PASS, Skill Scanner: PASS. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.
This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.
Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.
tomevault-io (@tomevault-io) published this skill. Their other Agent Skills are listed on their SkillMD profile.