kaixinbaba

ekscoding

Use this when the user asks for coding task/progress planning, executing the next unchecked task, executing tasks until a target module is complete, generating human+agent acceptance docs, interactive guided validation walkthrough, archiving plan history, or generating deployment checklists. Triggers on commands like /createTasks, /doNextTask, /doTasksUntil, /validateResult, /helpValidate, /archiveHistory, /generateDeploymentChecklist, or when you says things like "拆分任务"、"执行下一个任务"、"执行到 Module X"、"生成验收文档"、"帮我验收"、"引导我验证"、"一步步测"、"归档任务历史"、"生成部署清单"、"我要上线了准备清单".

kaixinbaba 0 Updated 3w ago

Resources

6
GitHub

Install

npx skillscat add kaixinbaba/ekscoding

Install via the SkillsCat registry.

SKILL.md

Ekscoding Workflow Skill

标准化开发交付工作流 Skill,将需求拆分、任务执行、验收、归档、部署准备串联为可重复的命令。

Purpose

Standardize development delivery into repeatable command workflows:

  • /createTasks: From a requirement, create task plan + progress tracking doc
  • /doNextTask: Execute the first unchecked task, update progress, git commit once done
  • /doTasksUntil: Execute tasks sequentially from first unchecked until target module is fully complete (includes all prerequisite tasks)
  • /validateResult: Generate dual-track acceptance docs (human + agent) + run agent checks
  • /helpValidate: Interactive guided validation — walk through acceptance checklist step-by-step with Q&A, troubleshoot issues together, explain the "why" behind each test
  • /archiveHistory: Archive all files under docs/plans/ into docs/history/YYYY-MM-DD/, generate summary, then clear docs/plans/
  • /generateDeploymentChecklist: Generate deployment checklist, clearly categorizing: 1) Agent can do now (no extra permissions), 2) Agent can do with user's credentials/tokens, 3) Must be done by user (detailed steps with minimal mental burden)

Command Router

When user intent matches one of the commands, route immediately to the workflow file:

  1. /createTasks -> workflows/create-tasks.md
  2. /doNextTask -> workflows/do-next-task.md
  3. /doTasksUntil -> workflows/do-tasks-until.md
  4. /validateResult -> workflows/validate-result.md
  5. /helpValidate -> workflows/help-validate.md
  6. /archiveHistory -> workflows/archive-history.md
  7. /generateDeploymentChecklist -> workflows/generate-deployment-checklist.md

If the user does not type the slash command but intent is equivalent, still route to the matching workflow.

Global Rules

  • All plan docs live in docs/plans/; history in docs/history/
  • Task and progress docs always use sequential numbers (task1.md, progress1.md, task2.md, ...)
  • When in doubt, keep docs in the same format as the previous ones
  • Never execute multiple tasks in one run for /doNextTask
  • Always commit git changes after completing one task in /doNextTask
  • /doTasksUntil executes multiple tasks sequentially until target module is complete; each task gets its own commit
  • Always verify acceptance criteria before marking a task complete

Required Output Quality

  • Tasks must be broken down to 1-2 hour size
  • Acceptance criteria must be objectively verifiable
  • All status changes must include timestamp + brief comment
  • Docs must be human-readable first, machine-parseable second

File Naming Conventions

Doc Type Path Pattern
Task Split docs/plans/task{N}.md
Progress Tracking docs/plans/progress{N}.md
Human Acceptance docs/plans/acceptance-{FEATURE}.md
Agent Acceptance docs/plans/acceptance-{FEATURE}-agent.md
Archived History docs/history/YYYY-MM-DD/
Deployment Checklist docs/plans/deployment-checklist-{PROJECT}.md

State Markers (Universal)

Marker Meaning Usage
[ ] Not started Task/acceptance item pending
[x] Done Task completed and verified
[~] Blocked Stopped due to external issue; MUST include reason

Deployment Checklist Categories

  1. Agent can do now (no extra permissions)
  2. Agent can do with user's credentials/tokens
  3. Must be done by user — detailed step-by-step, minimize mental burden