Token导航 LogoToken导航TokenDH.com
研究检索只读github未标认证来源可访问许可证需确认审计通过

finalise-plan敲定计划

Agent Skill

finalise-plan 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

总安装

196

周安装

8

GitHub Stars

公开资料未说明

下载量

63
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

unknown

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:finalise-plan(敲定计划)
来源仓库:https://github.com/sebkay/skills
仓库路径:skills/finalise-plan
安装命令:
npx skills add https://github.com/sebkay/skills --skill finalise-plan
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/sebkay/skills --skill finalise-plan

简介

用于查找、检索和筛选相关信息。finalise-plan 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

  • 适合在关键词搜索或任务场景下快速定位候选结果。
  • 可结合来源仓库和原始 README 进一步核验具体用法。
  • 安装前建议确认权限范围和维护状态,避免意外操作。
  • 注意检查是否会触发联网、命令执行或文件读写行为。

SKILL.md

Finalise Plan

Review the current plan like a critical peer review. Find the real problems, then update the plan so the fixes are captured in the plan itself.

Commit Discipline

Treat per-task commits as an execution constraint, not a reporting preference.

  • Every implementation task in the plan MUST be small enough to be completed, validated, and committed independently.
  • Every implementation task MUST include:

- Commit: the exact commit boundary for that task - Validation: the checks that must pass before committing

  • The implementer MUST stop after each completed task, run that task’s validation, and create the commit before beginning the next task.
  • If work for multiple tasks accumulates in the same uncommitted worktree, that is a plan failure. Revise the plan into smaller tasks first.
  • Do not defer commits until the end of implementation.
  • If a task cannot realistically be committed independently, it is too large or incorrectly scoped and the plan MUST be revised before execution.

Workflow

  1. Read the full request, surrounding context, and current plan state before judging the plan.
  2. Look for holes:

- missing scope boundaries - missing discovery or dependency checks - missing implementation steps - missing validation or rollout steps - missing final reporting or follow-up work - missing commit checkpoints or unclear commit boundaries

  1. Look for mistakes:

- incorrect sequencing - steps that are too broad to execute safely - assumptions presented as facts - hidden blockers, prerequisites, or irreversible actions - acceptance criteria that are vague or untestable - tasks that cannot be validated and committed independently - plans that imply batching multiple tasks into one commit

  1. If anything material is still unclear, ask the user to clarify with the ask questions tool before finalizing the plan.
  2. Update the plan directly to reflect the findings:

- add missing steps - split broad steps into atomic work - reorder steps to respect dependencies - add explicit validation tasks - add explicit per-task commit checkpoints - document unresolved risks or assumptions when they cannot yet be removed

  1. After updating the plan, briefly summarize the key corrections and any remaining open risks.
  2. Revise the plan using the /planner skill.
  3. You MUST structure the plan so each task ends in its own validation checkpoint and atomic commit, and the implementer must commit immediately after completing that task.

Review Standard

  • Prefer fixing the plan over merely describing problems.
  • You MUST structure the plan so each task ends in its own validation checkpoint and atomic commit.
  • Treat tasks without an explicit validation-and-commit boundary as defective.
  • Treat deferred end-of-work batching commits as a sequencing error, not a style choice.
  • You MUST make an atomic commit for each task in the plan, and revise the plan if any task is too large or mixed to support that.
  • Be skeptical of plans that jump straight to implementation.
  • Treat missing validation as a real defect.
  • Treat unclear assumptions as defects unless they are explicitly called out.
  • Ask only clarifying questions that materially change the plan.
  • Stop once the plan is concrete, sequenced, and testable.

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

Codex

33.66%
按下载量换算21

Claude

31.33%
按下载量换算20

Cursor

20.38%
按下载量换算13

Gemini CLI

8.88%
按下载量换算6

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

只读

该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills