# Executing Plans

> Use when: execute a written implementation plan with checkpoints, verification, and blocker handling.

- Skill: `kimtth/executing-plans` (Agent Skill)
- Install (CLI): `npx skillmds@latest add kimtth/executing-plans`
- Raw SKILL.md: https://api.skillmd.com/api/skills/kimtth/executing-plans/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: kimtth (https://skillmd.com/u/kimtth)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/kimtth/executing-plans

---


Goal: turn an approved plan into completed, verified work.

Use for:
- plan files with ordered implementation steps
- coding sessions where scope is already agreed
- carrying work through tests, fixes, and final report

Workflow:
1. Load the plan and read it critically before editing.
2. Raise gaps, contradictions, or unsafe steps before starting.
3. Convert plan steps into a live task list.
4. Execute one task at a time and mark status as it changes.
5. Run the verification specified by the plan; add targeted checks if missing.
6. Stop on blockers instead of guessing through unclear instructions.
7. Finish with tests, diff summary, and remaining risk.

Stop when:
- a dependency is missing
- verification fails repeatedly
- instructions conflict with repo reality
- scope expands beyond the approved plan

Rules:
- do not skip plan review
- do not claim completion without fresh evidence
- keep edits scoped to the plan
- ask only when a blocker cannot be resolved locally