# Implementation Agent

> Turn a plan/PRD markdown file into an executable task-list.md, then implement tasks incrementally while updating the checklist.

- Skill: `shrubnetllc/implementation-agent` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add shrubnetllc/implementation-agent`
- Raw SKILL.md: https://api.skillmd.com/api/skills/shrubnetllc/implementation-agent/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Product & Planning
- Author: shrubnetllc (https://skillmd.com/u/shrubnetllc)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/shrubnetllc/implementation-agent

---


# Implementation Agent (plan → tasks → incremental delivery)

## Inputs
- Plan file: use the file the user names (e.g., @my-plan.md). If none is named, default to @crewai_consolidation.md.
- Checklist file: task-list.md (create if missing).

## Hard rules
- Treat the plan as authoritative for scope and ordering.
- Never implement outside the plan unless the user explicitly expands scope.
- Update `task-list.md` after each completed step chunk.
- When updating `task-list.md`, edit ONLY the bracket contents (keep step ranges as-is).

## Procedure
1. Read the plan file end-to-end.
2. If `task-list.md` does not exist:
   - Generate it from the plan as a flat list of tasks with stable IDs.
   - Each task line must include a fixed step range that maps back to the plan.
3. Select the next incomplete task from `task-list.md`.
4. Implement only what that task requires.
5. Run the smallest relevant checks/tests.
6. Update `task-list.md` brackets with completed step numbers.
7. Repeat until all tasks are complete.

## Output format requirements for task-list.md
- Start with a short "Instructions" header.
- Then a task list, one task per line:
  task-<major>.<minor>: [] (<short label>; steps X–Y)

## If the plan is ambiguous
- Make the smallest reasonable assumption and record it in a short note at the top of `task-list.md` under a "Notes" section.