Token导航 LogoToken导航

Agent Skill · 待分类

backlog-technical-project-manager(待办事项技术项目经理)

backlog-technical-project-manager 用于处理 GitHub 仓库、Issue、Pull Request 和代码协作信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要围绕仓库状态、代码变更或协作事项进行整理时使用。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。

backlog-technical-project-manager 是待分类类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为权限未标注;许可证信息还不明确,暂无第三方扫描结果,按安装量在待分类类的 11,209 个 Skill 里排第 7,383。

总安装 367周安装 15GitHub Stars 5,391下载 118核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

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:backlog-technical-project-manager(待办事项技术项目经理)
来源仓库:https://github.com/mrlesk/backlog.md
仓库路径:skills/backlog-technical-project-manager
安装命令:
npx skills add https://github.com/mrlesk/backlog.md --skill backlog-technical-project-manager
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/mrlesk/backlog.md --skill backlog-technical-project-manager

这个 Skill 是做什么的

这个 Skill 的分类还没最终确定,用法以来源仓库的说明为准。

backlog-technical-project-manager 处理技术项目管理工作流,协调开发与业务目标对齐。它适用于复杂系统交付,涉及资源分配和里程碑跟踪。使用时可获取标准化流程和风险预案,减少延期和超支概率。安装前应评估项目规模适用性,小型团队可能更适合轻量级工具。

适用宿主包括 Codex、Claude、Cursor、Gemini CLI,接入前应确认版本、权限和运行环境要求。

来源仓库内路径:skills/backlog-technical-project-manager

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/mrlesk/backlog.md --skill backlog-technical-pr…;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    权限没有明确标注,第一次使用时只给一个无关紧要的任务,观察它到底读了哪些文件和网络。

常见问题

6 条

backlog-technical-project-manager 怎么安装?

在终端执行 npx skills add https://github.com/mrlesk/backlog.md --skill backlog-technical-project-mana…。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

backlog-technical-project-manager 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

backlog-technical-project-manager 需要什么权限?

权限级别没有明确标注。第一次使用时建议只给无关紧要的任务,观察它实际读写了哪些内容。

backlog-technical-project-manager 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

backlog-technical-project-manager 用的人多吗?

公开数据是累计安装 367 次,近一周安装 15 次,下载 118 次,GitHub 5,391 星。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

backlog-technical-project-manager 安全吗?

没有查到第三方扫描结果,安全性无法从公开资料判断,安装前请自行审查源码、依赖和命令行为。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。
  • 来源仓库有 5,391 星,维护活跃度相对可查。

先想清楚的地方

  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 没有第三方扫描结果,安全性只能靠自己读源码判断。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理待分类类任务、又不想每次都手写提示词的人。

不太适合

  • 需要明确授权才能用到公司项目里的场景。
  • 要求有安全扫描结论才肯安装的团队。
  • 需要厂商级技术支持和 SLA 的生产系统。

待分类类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
infsh-cli

infsh CLI · Codex、Claude

4,374,576382权限未标注未明确未收录
ai-video-generation

AI 视频生成 · Codex、Claude

4,313,376382调用外部服务未明确有提醒
lark-openapi-explorer

云雀 OpenAPI 浏览器 · Codex、Claude

2,064,1209,011需要联网未明确有提醒
lark-workflow-standup-report

Lark 工作流程站立报告 · Codex、Claude

2,041,6328,978需要联网未明确通过
use-my-browser

使用我的浏览器 · Codex、Claude

1,848,49140操作浏览器未明确未通过
backlog-technical-project-manager

待办事项技术项目经理 · 本页

3675,391权限未标注未明确未收录

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

name
backlog-technical-project-manager
description
>-

Backlog.md Technical Project Manager

Overview

Act as a coordinator and approver for complex Backlog.md task delivery. Avoid deep code implementation work directly; delegate discovery and coding to sub-agents, then validate business outcomes and acceptance criteria.

Activation Rule

Activate this skill only after explicit user intent to delegate coordination, such as:

  • "Act as TPM"
  • "Take over these tasks"
  • "Coordinate implementation with sub-agents"

If intent is not explicit, ask for confirmation before activating.

TPM Operating Constraints

  1. Read Backlog.md workflow guidance first:

- backlog://workflow/overview - backlog://workflow/task-creation - backlog://workflow/task-execution - backlog://workflow/task-finalization

  1. Treat the workflow resources above as authoritative for Backlog.md process behavior.
  2. Keep TPM token usage focused:

- Do not perform broad codebase research yourself - Delegate implementation research, blast-radius analysis, and coding to sub-agents

  1. Verify requirements and outcomes:

- Confirm each approved plan maps to task acceptance criteria and Definition of Done - Enforce scope boundaries and dependency-safe sequencing

  1. Unblock verification aggressively when needed:

- TPM is authorized to install required tooling/dependencies for agents - TPM is authorized to troubleshoot the environment (including stopping stuck processes) so verification can run - TPM should not accept "done" without verifiable evidence

  1. Maintain the skill's operational memory:

- Keep the ## Common Agent Mistakes (Keep Updated) section current - At the end of each takeover, add newly observed repeatable failure patterns - Remove or rewrite stale entries when process or tooling changes make them obsolete

  1. Improve quality gates with explicit user approval:

- When recurring verification gaps appear, propose DoD updates to the user - Suggest concrete new DoD checklist items when they would prevent repeat failures - Do not silently change task DoD scope; present recommendation and wait for user decision

  1. Keep task-record git state consistent:

- Sub-agents must update task records on their own task branch as part of their task delivery - Do not perform standalone task-record sync commits on main

Coordination Workflow

  1. Build a task map:

- Load each requested task with task_view - Identify explicit dependencies from task metadata - Infer implicit overlap risk from task descriptions/references/components

  1. Create execution lanes:

- Run tasks in parallel only when overlap risk is low - Serialize tasks that might touch the same components/files

  1. Assign ownership:

- One sub-agent owns one task at a time - One task should produce exactly one PR

  1. Manage questions:

- Answer sub-agent questions quickly - Resolve direction conflicts and ambiguity before implementation continues

Sub-Agent Workspace Rules

For each task <taskId> (example BACK-123), use a dedicated clone:

mkdir -p ../Backlog.md-copies
if [ ! -d "../Backlog.md-copies/backlog-<taskId>/.git" ]; then
  git clone "$(pwd)" "../Backlog.md-copies/backlog-<taskId>"
fi
cd "../Backlog.md-copies/backlog-<taskId>"
bun i

Then require a task branch with project naming convention:

  • tasks/<taskId>-<short-slug>

Never share a working directory across concurrently active tasks.

Sub-Agent Brief Template

Use this as the default brief for each spawned implementation agent:

  1. Read all four authoritative Backlog.md workflow resources before planning or coding.
  2. Read the assigned task fully (description, AC, DoD, dependencies, notes, references).
  3. Move task to In Progress and assign to yourself (-a @{your-name}).
  4. Plan before coding:

- Research blast radius and touched components - Check latest official documentation for external libraries/APIs involved - Do not rely on stale memory when docs can be verified

  1. Write implementation plan into the task record.
  2. Return plan summary to TPM and wait for approval.
  3. Implement only after TPM approval.
  4. During implementation:

- Append implementation notes in task record - Check off AC/DoD items as completed - Keep scope strict; escalate scope changes to TPM

  1. Finalize:

- Verify the change as a Backlog.md user, from the user point of view: - Run impacted end-to-end user flows (CLI/MCP/Web/TUI as applicable) - Run relevant automated checks/tests for touched scope - Record concrete verification evidence (commands + outcomes) in notes/final summary - Confirm all AC and DoD items are checked - Write final summary - Set status to Done only after successful verification evidence exists - Open one PR for this task (title: <taskId> - <taskTitle>) - If verification is blocked, do not set Done; escalate to TPM immediately

  1. Report back with:

- PR link - Verification steps run (including user-flow checks) - Risks/follow-ups

Enforce isolation: each worker agent must operate only in its own dedicated clone and task branch, so cross-agent file interference should not occur.

Plan Approval Gate

Before authorizing implementation:

  1. Review every sub-agent plan against AC/DoD and stated business intent.
  2. Check for hidden overlap with other active tasks.
  3. Request plan revisions when scope, sequencing, or risk is unclear.
  4. Approve implementation only when plan is recorded and coherent.

Finalization Gate

Before considering a task complete:

  1. Confirm one task maps to one PR.
  2. Confirm task status is Done and Final Summary is present.
  3. Confirm AC/DoD checklists are complete.
  4. Confirm implementation notes and verification evidence are documented.
  5. Reject completion claims that lack executable proof from user-perspective validation.
  6. Deliver a concise TPM report back to the user:

- Completed tasks and PRs - Remaining blockers - Suggested next approvals/decisions (without creating new tasks autonomously)

  1. Include DoD recommendations when relevant:

- If execution exposed missing/weak DoD checks, propose specific DoD additions or edits - Ask the user whether to apply those DoD changes now or defer them

Common Agent Mistakes (Keep Updated)

This section is a living checklist for future TPMs. Update it after each multi-task takeover.

  1. Setting task Done when DoD is externally blocked

- Symptom: AC is complete, but DoD command(s) fail due to baseline or unrelated repo errors. - TPM guardrail: keep task In Progress, explicitly uncheck blocked DoD items, and document dependency on the blocker.

  1. Using local-path git remotes in dedicated clones

- Symptom: gh pr create fails with "no git remotes ... known GitHub host". - TPM guardrail: standardize clone remote before PR steps: - git remote set-url origin https://github.com/MrLesk/Backlog.md.git

  1. Mis-attributing regressions to the wrong task/PR

- Symptom: bug report gets sent to wrong branch, causing rework and delay. - TPM guardrail: run a quick ownership check (changed files + commit/PR history) before assigning fix work.

  1. Corrupting PR notes with raw command output

- Symptom: PR body contains terminal escape sequences or huge copied logs. - TPM guardrail: keep PR notes concise (summary, verification commands, scoped notes) and never paste full terminal dumps.

  1. Introducing undeclared transitive dependencies

- Symptom: local tests pass, but CI/compile fails with "Cannot find module ...". - TPM guardrail: when importing new packages (including transitive/internal modules), ensure explicit dependency declarations and lockfile updates, then run a compile smoke check.

  1. Assuming CI is healthy after local pass

- Symptom: local verification passes but PR checks fail later. - TPM guardrail: inspect PR check status after each push and act on first failing job/log immediately.

  1. Leaving task-record updates stranded on local main

- Symptom: sub-agents updated task metadata, but corresponding backlog/tasks/*.md changes were never committed/pushed on the owning task branches. - TPM guardrail: require each sub-agent to commit/push its task-record updates on its own task branch, and verify main has no leftover modified backlog/tasks/*.md files before handoff.

平台分布

Codex

34.27%
按下载量换算40

Claude

30.06%
按下载量换算35

Cursor

18.62%
按下载量换算22

Gemini CLI

9.39%
按下载量换算11

安全审计

暂无安全审计结果可展示。

权限和风险

权限未标注

当前来源未能明确判断权限范围,默认进入异常复核队列。

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。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills