AgentSkillsCN

executing-tasks

为各类任务提供可复用的六阶段任务执行工作流,涵盖上下文收集、计划制定、任务执行、质量检查、进度更新以及学习文档记录等环节。在从故事分解中实施任务、执行开发工作或运行结构化任务流程时,应优先使用此技能。支持“执行任务”、“任务工作流”、“实施任务”、“任务阶段”、“质量检查”、“运行任务”、“完成任务”等触发条件。

SKILL.md
--- frontmatter
name: executing-tasks
description: Provides reusable 6-phase task execution workflow for any task type including context gathering, planning, execution, quality check, progress update, and learning documentation. This skill should be used when implementing tasks from story breakdowns, executing development work, or running structured task workflows. Triggers on "execute task", "task workflow", "implement task", "task phases", "quality check", "run task", "complete task".

Task Execution

Standardized 6-phase workflow for executing any task type with consistent quality gates and progress tracking. Integrates with sprint-orchestration for task assignment and progress updates.

6-Phase Workflow

text
[1. CONTEXT] --> [2. PLANNING] --> [3. EXECUTION] --> [4. QUALITY] --> [5. PROGRESS] --> [6. LEARNING]
     |               |                  |                 |                |                 |
  Load skill    TodoWrite tasks    Execute work      Verify results   Update YAML      Document
  Gather info   15-30 min chunks   Atomic commits    Run checks       Comment issue    learnings

Phase Summary

PhasePurposeKey Actions
1. ContextGather task-specific informationLoad relevant skills, read specs/designs, identify dependencies
2. PlanningBreak work into trackable unitsTodoWrite with 15-30 min granularity, identify parallelizable work
3. ExecutionImplement the actual workFollow task-type patterns, atomic commits, delegate to subagents
4. QualityVerify correctnessRun verification commands, record results, handle errors
5. ProgressUpdate tracking systemsPROGRESS.yaml transitions, GitHub issue comments, metrics
6. LearningCapture improvementsDocument blockers, patterns discovered, process improvements

Completion Criteria by Task Type

Task TypeVerificationSuccess Criteria
Code implementationnpm run typecheck && npm run testAll tests pass, no type errors
Test writingnpm run test -- <test-file>Tests pass, coverage meets target
DocumentationFrontmatter validation, link checksAI-readable, accurate content
RefactoringTests pass, no behavior changeAll existing tests green
Bug fixRegression test added and passesIssue resolved, test prevents recurrence
ConfigurationManual verification or smoke testSystem functions correctly

Quality Gates

Before transitioning from Execution (Phase 3) to Quality (Phase 4):

  • All planned TodoWrite items completed or explicitly deferred
  • Atomic commits made for logical units of work
  • No uncommitted changes unless intentionally staged

Before transitioning from Quality (Phase 4) to Progress (Phase 5):

  • All verification commands executed
  • Any failures documented with root cause
  • Re-execution loop if quality checks fail

References

ReferencePurpose
references/phase-1-context.mdContext gathering strategies by task type
references/phase-2-planning.mdTodoWrite patterns and task granularity
references/phase-3-execution.mdExecution patterns and subagent delegation
references/phase-4-quality.mdQuality gates and error handling
references/phase-5-progress.mdPROGRESS.yaml updates and issue commenting
references/phase-6-learning.mdLearning documentation format
assets/task-prompt-template.mdRalph-loop prompt template for task execution