GoalPro skill
A skill that turns a vague request into a Goal Prompt with success criteria and stop conditions, plus a Loop Prompt if needed.
- 57 ★ on the repository
- Last commit
- MIT
Minimal access
Why this level
- The skill contains only instructions and runs or sends nothing.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/goalproDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- The skill contains only instructions and runs or sends nothing.
This is third-party code. Review the repository files before installing.
What it does
GoalPro writes a task prompt for an executor: Codex, Claude Code or another agent. The input is a vague or emotional request and the output is a Goal Prompt with a goal contract. It separates stated intent from inferred intent, marks value judgments a person must confirm, and describes the outcome, win and failure criteria, required evidence, exclusions and pause conditions. By default the skill outputs only the Goal Prompt and stops. A Loop Prompt is added when delivery will produce new evidence for the next step or the user asks for continuation. It opens with a field for the next run time and diagnoses gaps between goal and result. Each round ends with one outcome: done, continue or pause.
Who it is for. People who assign tasks to coding agents and lose time to clarifications, context sprawl and "done" claims without proof.
Good fit when
- A task is stated vaguely and the agent starts changing code without a plan.
- You need a pause and a question to a person before deleting data, publishing or touching secrets.
- Run results are expected after delivery and you need a next-round template.
- The executor must attach an example showing whether the goal was met.
Not a fit when
- You need an agent to execute the task: the skill only writes the prompt.
- The task is tiny and obvious and one line suffices.
- You need a scheduled background run: the time in a Loop Prompt creates no job by itself.
Example request
Write a high-quality goal so Codex fixes this project and stops drifting off task.Limitations
The repository moved into the Kim Service collection and gets no new standalone releases: look there for fresh versions and install instructions. The skill executes nothing and creates no background jobs even when a Loop Prompt names a time. The executor starts only after separate user permission. Documentation is mostly in Chinese. Goal Prompt quality depends on how precisely you stated the request, and the repository check `node --test tests/governance.test.mjs` only compares the two skill copies, not model behaviour.
How to disable. Delete the `goalpro` folder from `~/.claude/skills` or from the Codex skills directory.
FAQ
What does GoalPro output by default?
Only a Goal Prompt you can copy to an executor. The skill adds no commentary around it.
When does a Loop Prompt appear?
When the user asks for continued iteration or delivery will produce data that decides the next step.
What ends a round?
Each round ends with `Done`, `Continue` or `Pause`. Pause follows two rounds without convergence, an impossible verification or a needed human decision.
Where are the Claude Code and Codex files?
`.claude/skills/goalpro/` for Claude Code and `.agents/skills/goalpro/` for Codex. Both copies hold the same content.
Does it create a schedule?
No. The time in a Loop Prompt is a hint, and a real scheduled run is configured separately.
Discussion
No questions or reviews yet.
Related
SkillsA skill that writes precise prompts tailored to a specific tool: Cursor, Claude Code, Midjourney, Stable Diffusion and more
A skill that scores a prompt on five criteria, picks one of 31 frameworks and rewrites the request after clarifying questions
A skill that builds a Codex /goal command from a five-section template so a long task does not drift and passes the audit
RePrompter interviews you about a rough request and turns it into a scored, structured prompt or a build prompt for a project.
A meta-skill for Codex: a vague task becomes a ready "/goal" command with verification, boundaries and pause conditions.
A skill that turns a goal into a ready prompt: the agent works in builder and critic loops until its output wins a blind comparison with a reference
README in short
The README describes the goal contract, Goal Prompt and Loop Prompt formats, pause rules and copying the folder for personal or project installs. The repository holds two byte-identical copies of the skill.
README badge
Original SKILL.md
--- name: goalpro description: 当用户要写出高质量 Goal Prompt，把模糊、战略性、多步骤、证据不足、持续/自动化或容易跑偏的请求整理成可执行、可验证、可暂停的目标契约时使用。适用于写 goal、优化任务提示词、明确 done/success criteria、deep research 后定战略、大改前 inventory、修复跑偏计划、识别可委托的重复 workflow、为 Codex 或 Claude Code 准备执行任务；默认一次性交付只生成 Goal Prompt，只有存在真实后续迭代需求时才追加 Loop Prompt，不执行 goal 或 loop，也不创建自动化。 --- # Goal Prompt 目标：先区分用户明确表达的意图、AI 推断的潜在意图和需要用户确认的价值判断，再把战略判断、证据标准和成败边界写成 Codex / Claude Code 能执行、能验收、少跑偏的 Goal Prompt。只有存在真实后续迭代需求时，才追加交付后使用的 Loop Prompt。 GoalPro 的交付物是可复制的提示词，不是任务执行结果。默认一次性交付只输出 `Goal Prompt`。只有用户明确要求持续迭代 / LOOP，或任务在本次交付后会持续产生新证据并需要下一轮动作时，才追加 `Loop Prompt`。Loop 不是“一次性返工提示词”，而是每轮都产出 `Next LOOP packet` 的循环控制器；它必须在开头给出 `时间参数`，让用户填写下一轮什么时候继续。除非用户明确说“按这个 goal 执行 / 开始改 / 写入文件 / 提交 / 创建自动化”，否则输出适用的提示词后必须停止。 GoalPro 不选择或路由组件，不协调跨组件状态机，也不执行目标、运行 Loop 或替用户做最终验收；后续动作必须由另行获授权的执行者承担。 这不是“让提示词更短”的 Skill。表达经济只在战略完整后处理：删空话，不删判断、边界、证据和验收。 ## 先判任务级别 - `Intake`：用户只要更好的 goal / prompt / spec。 - `Strategic`：用户要战略、方案、路线、标准、重要决策或高质量研究结论。 - `Execution`：用户要给 agent 一份按 goal 开始做的执行提示词。 - `Repair`：之前输出跑偏、太粗糙、太复杂、问太多、假完成。 - `Governed`：高风险、多文件、发布、外部事实、生产相邻或会影响真实用户的任务。 - `Workflow`：用户提到每天、每周、自动、持续、发布、运营、监控、队列、复盘等重复性工作时，先判断哪些 Trigger / Checkpoint / Brief 信息应写进 Goal Prompt；只有通过 Loop 需求判断门才追加 Loop，开发一个自动化产品本身不等于授权交付后持续执行。 用能诚实验收的最轻模式；但战略性任务必须先过证据门槛。 ## Prompt-only 闸门 - Skill mention 不等于执行授权。用户只说 `goalpro`、`写 goal`、`优化提示词`、`帮我做个目标` 时，只生成可复制的 goal 提示词。 - 默认一次性交付只生成一段 `Goal Prompt`，不为了显得完整而附送 Loop。 - 只有通过“Loop 需求判断门”时才追加 `Loop Prompt`。Loop 只是交付后可粘贴的继续进化提示词，不授权当前回合执行。 - Workflow lens 不新增第三个 `Workflow Prompt`，不把 GoalPro 变成执行器或自动化平台；必要的 workflow 信息写进 Goal，存在真实循环时再写进 Loop。 - Loop Prompt 必须把 `时间参数` 放在最前面：提示用户自行填写 LOOP 时间，如“手动：贴入上一轮结果后继续”或“每天早上 09:00”。 - 定时/自动 Loop 只是自动化设置说明，除非用户明确授权创建或修改自动化；写了时间不等于已经创建后台任务。 - 生成的 goal 必须能指导后续执行者：对象、动作、上下文、范围、非目标、检查点、暂停条件和验收证据都要清楚。 - 生成的 loop 必须能指导后续执行者复盘上一轮真实结果：看最终报告、diff、验证、截图、用户反馈，定位剩余差距，再决定本轮动作、是否已 Done、是否 Pause、按哪个时间参数继续、以及下一轮 `Next LOOP packet`。 - 不要因为 goal 里写了 Execution policy、Verification、Checkpoints，就在当前回合继续执行这些内容。 - 不要因为 loop 里写了 Review evidence、Cycle action、Verification delta，就在当前回合开始复盘或修复。 - 只有用户额外明确授权执行、保存、修改文件、提交或发布时，才进入独立的执行任务；那已经不是 GoalPro 的默认输出模式。