How do I install the Plan Goal skill?

Run npx skillmds@latest add djalmajr/plan-goal 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.

Is Plan Goal safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Plan Goal?

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.

Is Plan Goal free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Plan Goal?

djalmajr (@djalmajr) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Plan Goal by djalmajr — Agent Skill

Name: Plan Goal
Author: djalmajr

Plan Goal

Write one implementation plan (with a sequential checklist) under .agents/plans/, copy that plan to the clipboard, then give a paste-ready prompt to execute the goal. Use when the user runs /plan-goal, or asks for "plan and goal", "prepare the goal", or paste-ready goal text (also PT: "plano e goal", "prepara o goal").

djalmajr 0a8b672 2 files · 3.5 KB Updated Sep 17, 2026

File contents

/plan-goal

Write one plan file, copy it to the clipboard, and return paste-ready goal text. Do not implement. Do not start execution.

$ARGUMENTS is the objective. If it is empty, ask in one sentence what to plan and wait.

Project root

Write artifacts in the repo where the work will happen, not the agent's current working directory if they differ. Paths below are relative to that project root. If the root is ambiguous, ask before writing.

Files

One markdown, named from a stable slug of the objective: .agents/plans/<slug>.md

Create .agents/plans/ if it is missing. If .agents/plans/ is not yet in the repo .gitignore, add that line (do not ignore the rest of .agents/). Do not commit the plan unless the user asks.

The plan includes, in this order:

Context, decisions, AS-IS/TO-BE, files, phases, tests.
Goal kind, numbered Acceptance criteria, Verification plan (gating/evidence), Non-goals, Assumed scope, Implementation approach.
Task checklist — the progress machine.
Risks.

Do not write a second *.goal.md.

Clipboard

After writing the plan, copy that file's contents to the system clipboard and say that it is there. Prefer, in order:

macOS: pbcopy < .agents/plans/<slug>.md
Linux: wl-copy < .agents/plans/<slug>.md or xclip -selection clipboard
Windows: clip < .agents/plans/<slug>.md (or Get-Content … | Set-Clipboard)

If the copy fails, say so. Still print the closing fence.

Prompting

Empty objective: ask in free-form text (one sentence) and wait. Do not use the structured question tool — the objective is free-form.

Content

Language of the generated plan: the user's language. Code identifiers in English.
Acceptance criteria the agent can verify locally (tests, curl, an already-logged-in browser). Do not depend on a human click if it can be avoided.
Git only what this conversation authorized. No merge to main unless asked.
One goal per objective. Phases and cycles are Task checklist items (- [ ]), in order, each with its done-when in the item itself. The implementer marks [x] and continues without waiting and without a new goal. The first open box is the next step; the goal is reached only when the checklist and the gates are [x].
Each checklist item touches at most 5 code files (locales, CHANGELOG, and generated migrations extra). That limits the step, not the number of goals. Implementation approach must walk the checklist in that order.

Required closing

One fence, ready to paste into any agent, citing the plan path, the branch, the deliverable, the non-goals, and acceptance. Do not ask the user to rewrite the objective. The fence is a self-contained prompt. If the product has a goal runner (/goal or equivalent), the user may prefix it — whoever ran /plan-goal does not start that run.

---name: plan-goaldescription: Write one implementation plan (with a sequential checklist) under .agents/plans/, copy that plan to the clipboard, then give a paste-ready prompt to execute the goal. Use when the user runs /plan-goal, or asks for "plan and goal", "prepare the goal", or paste-ready goal text (also PT: "plano e goal", "prepara o goal").---# /plan-goalWrite **one** plan file, copy it to the clipboard, and return paste-ready goaltext. Do not implement. Do not start execution.`$ARGUMENTS` is the objective. If it is empty, ask in one sentence what toplan and wait.## Project rootWrite artifacts in the repo where the work will happen, not the agent's currentworking directory if they differ. Paths below are relative to that project root.If the root is ambiguous, ask before writing.## FilesOne markdown, named from a stable slug of the objective:`.agents/plans/<slug>.md`Create `.agents/plans/` if it is missing. If `.agents/plans/` is not yet in therepo `.gitignore`, add that line (do not ignore the rest of `.agents/`). Do notcommit the plan unless the user asks.The plan includes, in this order:1. Context, decisions, AS-IS/TO-BE, files, phases, tests.2. `Goal kind`, numbered `Acceptance criteria`, `Verification plan` (gating/evidence), `Non-goals`, `Assumed scope`, `Implementation approach`.3. `Task checklist` — the progress machine.4. `Risks`.Do not write a second `*.goal.md`.## ClipboardAfter writing the plan, copy **that file's contents** to the system clipboardand say that it is there. Prefer, in order:- macOS: `pbcopy < .agents/plans/<slug>.md`- Linux: `wl-copy < .agents/plans/<slug>.md` or `xclip -selection clipboard`- Windows: `clip < .agents/plans/<slug>.md` (or `Get-Content … | Set-Clipboard`)If the copy fails, say so. Still print the closing fence.## Prompting- Empty objective: ask in free-form text (one sentence) and wait. Do not use the structured question tool — the objective is free-form.## Content- Language of the generated plan: the user's language. Code identifiers in English.- Acceptance criteria the agent can verify locally (tests, curl, an already-logged-in browser). Do not depend on a human click if it can be avoided.- Git only what this conversation authorized. No merge to `main` unless asked.- **One goal per objective.** Phases and cycles are `Task checklist` items (`- [ ]`), in order, each with its done-when in the item itself. The implementer marks `[x]` and continues without waiting and without a new goal. The first open box is the next step; the goal is reached only when the checklist and the gates are `[x]`.- Each checklist item touches at most 5 code files (locales, CHANGELOG, and generated migrations extra). That limits the step, not the number of goals. `Implementation approach` must walk the checklist in that order.## Required closingOne fence, ready to paste into any agent, citing **the plan path**, the branch,the deliverable, the non-goals, and acceptance. Do not ask the user to rewritethe objective. The fence is a self-contained prompt. If the product has a goalrunner (`/goal` or equivalent), the user may prefix it — whoever ran`/plan-goal` does not start that run.

Frequently asked questions