Guide · Agent skill
How to use the Execute Plan agent skill (execute-plan)
Updated · written with AI from the project's own documentation and checked against our catalogue data
The execute-plan skill is an agent skill that tells an AI agent how to carry out a structured task plan from start to finish, marking each task as it goes. It is aimed at developers and teams who keep multi-step plans in markdown files. After this guide you will know what the skill changes about the agent's behaviour, how to install it, how to trigger it with a request, and what to watch for before running it on a real plan.
Execute Plantailcallhq/forgecode · Apache-2.0 Download the skill Quick start
- 1Download the skill. Download the execute-plan ZIP from this skill's page in the directory.
- 2Unzip into a skills folder. Unzip the archive into ~/.claude/skills/ for personal use, or into .claude/skills/ in a project.
- 3Or upload it in the Claude apps. In the Claude apps, go to Settings → Capabilities → Skills and upload the skill.
- 4Write or check your plan file. Create a markdown checklist in the format plans/{current-date}-{task-name}-{version}.md with each task marked [ ].
- 5Ask the agent to execute it. Send a request that names the plan file, such as 'Execute plan at plans/2025-11-23-refactor-auth-v1.md'.
- 6Review the file afterwards. Check the status markers, including any [!] FAILED tasks, to see what was completed.
What the execute-plan skill makes the agent do
When you give the agent a plan file, this skill tells it to read the whole plan, work through the pending tasks one at a time, and record progress in the file. Each task is marked in progress when started and done when finished. At the end the agent is told to re-read the file to check that every task is complete before saying the plan is finished.
The skill is a workflow discipline, not a tool that does the work itself. The results depend on what the agent actually does for each task.
- Reads the whole plan file to find pending tasks
- Works through tasks one at a time, in order
- Marks each task IN_PROGRESS, then DONE, in the plan file
- Re-reads the plan at the end to verify completion
How to install the execute-plan skill
You can install the skill in one of two ways. Download the ZIP for execute-plan from the skill's page in this directory and unzip it into your personal skills folder, ~/.claude/skills/, or into a project's .claude/skills/ folder if you want it only for that project.
Alternatively, upload the skill in the Claude apps under Settings → Capabilities → Skills.
- Personal install: unzip into ~/.claude/skills/
- Project install: unzip into .claude/skills/ inside your project
- Claude apps: Settings → Capabilities → Skills, then upload
How to trigger the skill with a natural request
The skill is meant to activate when you provide a plan file path in the format plans/{current-date}-{task-name}-{version}.md, or when you explicitly ask the agent to execute a plan file. You do not need special syntax; a plain request that names the file is enough.
Plans are expected to be markdown checklists where each task has a status marker. The skill uses four markers: pending, in progress, done and failed.
- Execute plan at plans/2025-11-23-refactor-auth-v1.md
- Please execute the plan file plans/2025-11-23-refactor-auth-v1.md
- Work through the tasks in this plan file and track status as you go
[ ]: PENDING
[~]: IN_PROGRESS
[x]: DONE
[!]: FAILEDWhat the agent does step by step
The skill defines six steps. Before starting, the agent states a commitment to complete all tasks in the plan. It then reads the plan, announces the next pending task, marks it in progress, carries out the task, marks it done, and repeats until none are pending. A final re-read confirms completion before the agent announces it.
- Recite the commitment to complete all tasks
- Read the plan and identify pending tasks
- Announce the next task and mark it IN_PROGRESS
- Execute the task and mark it DONE
- Repeat until all tasks are DONE
- Re-read the plan to verify, then announce completion
What you need before you start
The skill lists no special requirements. It does not run scripts and does not use the network. What it does need is a plan that is well formed: a markdown file with tasks that have clear status markers and tasks specific enough to finish without a person making decisions along the way.
- A markdown plan file with checklist-style tasks
- Tasks that can be completed autonomously
- Status markers in the format shown above
Safety notes and when not to use it
The skill pushes the agent to finish every task without stopping for approval. For plans that include destructive or irreversible steps, such as deleting data or deploying to production, that means each step runs without a review. Keep those steps out of an automated run or review the plan carefully before you start.
Do not use the skill for quick one-off questions or small edits, where the plan overhead adds nothing. Plans with vague tasks, or tasks that depend on decisions only a person can make, are also a poor fit because the skill assumes each task can be completed autonomously.
- Avoid plans with deletions, data loss or production deploys unless you have reviewed them
- Do not rely on it for vague tasks or decisions that need a person
- Skip it for small one-off edits
Tips for getting good results from execute-plan
Write each task so it has a clear finish line. Keep the file name in the documented pattern so the trigger matches. Check the file after the run: the status markers show which tasks were started, finished or failed, which gives you a record of the run.
- Use the plans/{date}-{task}-{version}.md naming pattern
- Make each task concrete and checkable
- Review the status markers afterwards, including any FAILED entries
What people use it for
- Working through a multi-step refactoring plan kept in a markdown file
- Following a feature plan task by task with progress visible in the file
- Keeping a written record of which tasks were started, finished or failed
Frequently asked questions
Is the execute-plan skill free?
This page does not state a licence price. The source is published under the Apache-2.0 licence, which is listed for the item. Check the licence text in the repository for the terms that apply.
What do I need to install execute-plan?
The skill lists no special requirements, runs no scripts and uses no network access. You need to unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder, or upload it under Settings → Capabilities → Skills in the Claude apps.
How do I trigger the execute-plan skill?
Give the agent a plan file path in the format plans/{current-date}-{task-name}-{version}.md, or explicitly ask it to execute a plan file. A request such as 'Execute plan at plans/2025-11-23-refactor-auth-v1.md' is enough.
Does the skill run code or access the internet?
No. The skill is a SKILL.md instruction file, and the listing says it runs no scripts and uses no network. It shapes how the agent works through the plan, but the agent's actions on each task still depend on its own tools and your setup.
Can execute-plan run destructive tasks safely?
Not without review. The skill tells the agent to finish all tasks without stopping for approval, so plans with deletions, data loss or production deploys run unattended. Review those steps before starting, or remove them from the plan.
How is this different from just asking the agent to do a plan?
The skill adds a fixed workflow: a commitment statement, one task at a time, status markers written into the file, and a final re-read. Without it, the agent may not update the file or verify completion in the same way.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use execute-plan if you keep clear, checklist-style markdown plans and want the agent to follow them in order with progress recorded in the file. Pick something else for quick one-off edits, vague tasks that need human judgement, or plans with destructive steps you do not want run without review.
Download the skill