任务实现
ln-401-task-executor
by levnikolaevich
聚焦单项实现或重构,按任务说明推进 Todo→In Progress→To Review,更新 Linear/kanban 状态并执行 typecheck/lint,按需改文档和配置,不处理测试任务。
把开发任务从待办稳步推进到可评审阶段,按 KISS/YAGNI 落地并自带质量检查,适合想减少返工的团队。
安装
claude skill add --url github.com/levnikolaevich/claude-code-skills/tree/master/ln-401-task-executor文档
Paths: File paths (
shared/,references/,../ln-*) are relative to skills repo root. If not found at CWD, locate this SKILL.md directory and go up one level for repo root.
Implementation Task Executor
Executes a single implementation (or refactor) task from Todo to To Review using the task description and linked guides.
Purpose & Scope
- Handle one selected task only; never touch other tasks.
- Follow task Technical Approach/plan/AC; apply KISS/YAGNI and guide patterns.
- Update Linear/kanban for this task: Todo -> In Progress -> To Review.
- Run typecheck/lint; update docs/tests/config per task instructions.
- Not for test tasks (label "tests" goes to ln-404-test-executor).
Inputs
| Input | Required | Source | Description |
|---|---|---|---|
taskId | Yes | args, parent Story, kanban, user | Task to execute |
Resolution: Per shared/references/input_resolution_pattern.md — Task Resolution Chain.
Status filter: Todo
Task Storage Mode
MANDATORY READ: Load shared/references/tools_config_guide.md, shared/references/storage_mode_detection.md, and shared/references/input_resolution_pattern.md
Read docs/tools_config.md (bootstrap if missing per tools_config_guide.md).
Extract: task_provider = Task Management → Provider (linear | file).
| Aspect | Linear Mode | File Mode |
|---|---|---|
| Load task | get_issue(task_id) | Read("docs/tasks/epics/.../tasks/T{NNN}-*.md") |
| Update status | save_issue(id, state) | Edit the **Status:** line in file |
| Kanban | Updated by Linear sync | Must update kanban_board.md manually |
File Mode status format:
## Status
**Status:** In Progress | **Priority:** High | **Estimate:** 4h
Mode Detection
Detect operating mode at startup:
Plan Mode Active:
- Steps 1-2: Load task context (read-only, OK in plan mode)
- Generate EXECUTION PLAN (files to create/modify, approach) → write to plan file
- Call ExitPlanMode → STOP. Do NOT implement.
- Steps 3-6: After approval → execute implementation
Normal Mode:
- Steps 1-6: Standard workflow without stopping
Progress Tracking with TodoWrite
When operating in any mode, skill MUST create detailed todo checklist tracking ALL steps.
Rules:
- Create todos IMMEDIATELY before Step 1
- Each workflow step = separate todo item; implementation step gets sub-items
- Mark
in_progressbefore starting step,completedafter finishing
Todo Template (10 items):
Step 1: Resolve taskId
- Resolve via args / Story context / kanban / AskUserQuestion (Todo filter)
Step 2: Load Context
- Fetch full task description + linked guides/manuals/ADRs
Step 2b: Goal Articulation Gate
- Complete 4 questions from shared/references/goal_articulation_gate.md (<=25 tokens each)
Step 3: Start Work
- Set task to In Progress, update kanban
Step 4: Implement
- 4a Pattern Reuse: IF creating new file/utility, Grep src/ for existing similar patterns
(error handlers, validators, HTTP wrappers, config loaders). Reuse if found.
- 4b Follow task plan/AC, apply KISS/YAGNI
- 4c Architecture Guard: IF creating service function: (1) 3+ side-effect categories → split;
(2) get_*/find_*/check_* naming → verify no hidden writes; (3) 3+ service imports → flatten
- Update docs and existing tests if impacted
- Execute verify: methods from task AC (test/command/inspect)
Step 5: Quality
- Run typecheck and lint (or project equivalents)
Step 6: Finish
- Set task to To Review, update kanban
- Add summary comment (changes, tests, docs)
Workflow (concise)
- Resolve taskId (per input_resolution_pattern.md):
- IF args provided → use args
- ELSE IF Story context available → list Todo tasks under Story, suggest if 1
- ELSE IF kanban has exactly 1 Task in [Todo] → suggest
- ELSE → AskUserQuestion: show Todo Tasks from kanban
- Load context: Fetch full task description (Linear: get_issue; File: Read task file); read linked guides/manuals/ADRs/research; auto-discover team/config if needed.
2b) Goal gate: MANDATORY READ:
shared/references/goal_articulation_gate.md— Complete the 4-question gate (<=25 tokens each). State REAL GOAL (deliverable as subject), DONE LOOKS LIKE, NOT THE GOAL, INVARIANTS & HIDDEN CONSTRAINTS. - Start work: Update this task to In Progress (Linear: update_issue; File: Edit status line); move it in kanban (keep Epic/Story indent).
- Implement (with verification loop): Before writing new utilities/handlers, Grep
src/for existing patterns (error handling, validation, config access). Reuse if found; if not reusable, document rationale in code comment. Follow checkboxes/plan; keep it simple; avoid hardcoded values; reuse existing components; add Task ID comment (// See PROJ-123) to new code blocks; update docs noted in Affected Components; update existing tests if impacted (no new tests here). Before creating service functions, apply Architecture Guard (cascade depth, interface honesty, flat orchestration). After implementation, executeverify:methods from task AC: test → run specified test; command → execute and check output; inspect → verify file/content exists. If any verify fails → fix before proceeding. - Quality: Run typecheck and lint (or project equivalents); ensure instructions in Existing Code Impact are addressed.
- Finish: Mark task To Review (Linear: update_issue; File: Edit status line); update kanban to To Review; add summary comment (what changed, tests run, docs touched).
Pre-Submission Checklist
Context: Self-assessment before To Review reduces review round-trips and catches obvious issues early.
Before setting To Review, verify all 6 items:
| # | Check | Verify |
|---|---|---|
| 0 | AC verified | Each AC verify: method executed with pass evidence |
| 1 | Approach alignment | Implementation matches Story Technical Approach |
| 2 | Clean code | No dead code, no backward-compat shims, unused imports removed |
| 3 | Config hygiene | No hardcoded creds/URLs/magic numbers |
| 4 | Docs updated | Affected Components docs reflect changes |
| 5 | Tests pass | Existing tests still pass after changes |
| 6 | Pattern reuse | New utilities checked against existing codebase; no duplicate patterns introduced |
| 7 | Architecture guard | Cascade depth <= 2; no hidden writes in read-named functions; no service chains >= 3 |
If any check fails: Fix before setting To Review. Do not rely on reviewer to catch preventable issues.
Critical Rules
- Single-task updates only; no bulk status changes.
- Keep language of the task (EN/RU) in edits/comments.
- No code snippets in the description; code lives in repo, not in Linear.
- No new test creation; only update existing tests if required.
- Preserve Foundation-First ordering from orchestrator; do not reorder tasks.
- Add Task ID comments to new code blocks for traceability (
// See PROJ-123or# See PROJ-123). - Do NOT commit. Leave all changes uncommitted — ln-402 reviews and commits with task ID reference.
Definition of Done
- Task selected and set to In Progress; kanban updated accordingly.
- Guides/manuals/ADRs/research read; approach aligned with task Technical Approach.
- Implementation completed per plan/AC; each AC
verify:method executed with pass evidence. - Docs and impacted tests updated.
- Typecheck and lint passed (or project quality commands) with evidence in comment.
- Task set to To Review; kanban moved to To Review; summary comment added.
Reference Files
- Tools config:
shared/references/tools_config_guide.md - Storage mode operations:
shared/references/storage_mode_detection.md - Guides/manuals/ADRs/research:
docs/guides/,docs/manuals/,docs/adrs/,docs/research/ - Kanban format:
docs/tasks/kanban_board.md
Version: 3.0.0 Last Updated: 2025-12-23
相关 Skills
PPT处理
by anthropics
处理 .pptx 全流程:创建演示文稿、提取和解析幻灯片内容、批量修改现有文件,支持模板套用、合并拆分、备注评论与版式调整。
✎ 涉及PPTX的创建、解析、修改到合并拆分都能一站搞定,连备注、模板和评论也能处理,做演示文稿特别省心。
技能工坊
by anthropics
覆盖 Skill 从创建到迭代优化全流程:起草能力、补测试提示、跑评测与基准方差分析,并持续改写内容和描述,提升效果与触发准确率。
✎ 技能工坊把技能从创建、迭代到评测串成闭环,方差分析加描述优化,特别适合把触发准确率打磨得更稳。
Word文档
by anthropics
覆盖Word/.docx文档的创建、读取、编辑与重排,适合生成报告、备忘录、信函和模板,也能处理目录、页眉页脚、页码、图片替换、查找替换、修订批注及内容提取整理。
✎ 搞定 .docx 的创建、改写与精排版,目录、批量替换、批注修订和图片更新都能自动化,做正式文档尤其省心。
相关 MCP 服务
文件系统
编辑精选by Anthropic
Filesystem 是 MCP 官方参考服务器,让 LLM 安全读写本地文件系统。
✎ 这个服务器解决了让 Claude 直接操作本地文件的痛点,比如自动整理文档或生成代码文件。适合需要自动化文件处理的开发者,但注意它只是参考实现,生产环境需自行加固安全。
by wonderwhy-er
Desktop Commander 是让 AI 直接执行终端命令、管理文件和进程的 MCP 服务器。
✎ 这工具解决了 AI 无法直接操作本地环境的痛点,适合需要自动化脚本调试或文件批量处理的开发者。它能让你用自然语言指挥终端,但权限控制需谨慎,毕竟让 AI 执行 rm -rf 可不是闹着玩的。
EdgarTools
编辑精选by dgunning
EdgarTools 是无需 API 密钥即可解析 SEC EDGAR 财报的开源 Python 库。
✎ 这个工具解决了金融数据获取的痛点——直接让 AI 读取结构化财报,比如让 Claude 分析苹果的 10-K 文件。适合量化分析师或金融开发者快速构建数据管道。但注意,它依赖 SEC 网站稳定性,高峰期可能延迟。