Token导航 LogoToken导航

Agent Skill · 研究检索

dev-checkpoint(开发检查点)

dev-checkpoint 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。

dev-checkpoint 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为执行命令;许可证信息还不明确,第三方扫描已通过,按安装量在研究检索类的 73,675 个 Skill 里排第 62,244。

总安装 339周安装 14GitHub Stars 2下载 111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:dev-checkpoint(开发检查点)
来源仓库:https://github.com/andreaserradev-gbj/dev-workflow
仓库路径:skills/dev-checkpoint
安装命令:
npx skills add https://github.com/andreaserradev-gbj/dev-workflow --skill dev-checkpoint
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/andreaserradev-gbj/dev-workflow --skill dev-checkpoint

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

用于查找、检索和筛选相关信息,适合快速定位候选结果。可结合关键词、任务场景或来源线索进行信息聚合。通过命令行安装并使用,需参考原始 README 获取具体指令。安装前建议确认是否会触发联网或文件读写,确保权限可控。dev-checkpoint 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

来源仓库内路径:skills/dev-checkpoint

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/andreaserradev-gbj/dev-workflow --skill dev-ch…;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

    它需要执行命令,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

dev-checkpoint 怎么安装?

在终端执行 npx skills add https://github.com/andreaserradev-gbj/dev-workflow --skill dev-checkpoint。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

dev-checkpoint 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。

dev-checkpoint 需要什么权限?

权限级别是执行命令。安装流程涉及命令执行,可能通过 npx skills add https://github.com/andreaserradev-gbj/dev-workflow --skill dev-checkpoint 联网下载 Skill 或依赖。用户安装前应确认命令来源、仓库内容和执行环境。

dev-checkpoint 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。

dev-checkpoint 用的人多吗?

公开数据是累计安装 339 次,近一周安装 14 次,下载 111 次,GitHub 2 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。

dev-checkpoint 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

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。
  • 第三方扫描已通过,常见的高风险行为没有命中。

先想清楚的地方

  • 需要执行命令,装到常用环境前要先评估影响面。
  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 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。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。

不太适合

  • 不能接受 Skill 在你机器上执行命令的人。
  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
dev-checkpoint

开发检查点 · 本页

3392执行命令未明确通过

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

Checkpoint Current Session

Review the current session and create a continuation prompt for the next session.

SAVE-ONLY MODE

This skill analyzes and saves. It does NOT fix, investigate, or implement anything.

  • Do NOT investigate bugs or errors mentioned during the session
  • Do NOT start implementing fixes or next steps
  • Do NOT move to the next phase or task
  • If the user mentions bugs during confirmation (Step 6), note them in <blockers> or <notes> but do NOT attempt to fix them

Step 0: Discover Project Root

Run the discovery script:

bash "$DISCOVER" root

Where $DISCOVER is the absolute path to scripts/discover.sh within this skill's directory.

Path safety — shell state does not persist between tool calls, so you must provide full script paths on each call:

  • Use $HOME instead of the literal home directory (e.g., bash "$HOME/code/…/discover.sh", not bash "/Users/name/…/discover.sh"). This prevents username hallucination.
  • Copy values from tool output. When reusing a value returned by a previous command (like $PROJECT_ROOT), copy it verbatim from that command's output. Never retype a path from memory.
  • Verify on first call: if a script call fails with "No such file", the path is wrong — STOP and re-derive from the skill-loading context.
  • Never ignore a non-zero exit. If any script in this skill fails, stop and report the error before continuing.

Store the output as $PROJECT_ROOT. If the command fails, inform the user and stop.

Step 1: Identify the Active Feature

Run the discovery script to find features:

bash "$DISCOVER" features "$PROJECT_ROOT" "$ARGUMENTS"

Pass $ARGUMENTS as the third argument only if the user provided one; omit it otherwise.

  • If the script exits non-zero (no .dev/ directory): ask the user to specify the feature name.
  • If output is empty: no features found, ask the user to specify the feature name.
  • If one line: use as $FEATURE_PATH.
  • If multiple lines: ask which feature to checkpoint.

Never use raw $ARGUMENTS directly in shell commands or paths.

Validate with the validation script. Where $VALIDATE is the absolute path to scripts/validate.sh within this skill's directory. Apply the path safety rules from Step 0 ($HOME, copy from output).

If using an existing feature (a $FEATURE_PATH was matched):

bash "$VALIDATE" feature-path "$FEATURE_PATH" "$PROJECT_ROOT"

If creating a new feature (no match, normalizing user input):

bash "$VALIDATE" normalize "$USER_INPUT"

Outputs $FEATURE_NAME on success; on failure, STOP and report the error.

The checkpoint will be saved to $PROJECT_ROOT/.dev/$FEATURE_NAME/checkpoint.md.

Step 2: Analyze Session with Agent

Launch the checkpoint-analyzer agent to scan PRD files and the current session:

"Analyze the PRD files in $PROJECT_ROOT/.dev/$FEATURE_NAME/ and the current session.
Find: completed items (⬜ → ✅), pending items, decisions made, blockers encountered.
Determine current phase and next step."

Use subagent_type=dev-workflow:checkpoint-analyzer and model=haiku.

Step 3: Review Agent Findings

After the agent returns:

  1. Verify accuracy — Check that completed/pending items match what happened
  2. Add missing context — Include any decisions or blockers the agent missed

Step 4: Update PRD Status Markers (REQUIRED)

For each PRD file in .dev/$FEATURE_NAME/:

  1. Read the file
  2. Change ⬜ to ✅ for completed items; update "Status" fields
  3. Save changes

Track what was updated (file + markers changed) — reported in Step 9.

If nothing was completed, state: "No PRD updates needed."

Step 5: Capture Git State

Run the git state script:

bash "$GIT_STATE" full

Where $GIT_STATE is the absolute path to scripts/git-state.sh within this skill's directory. Apply the path safety rules from Step 0 ($HOME, copy from output).

Parse the output lines:

  • git:false → not a git repo; omit branch, last_commit, uncommitted_changes from frontmatter.
  • branch:<name> → store for frontmatter
  • commit:<oneline> → store as last commit
  • status:<line> → each is one line of git status --short; if no status: lines, working tree is clean

Step 6: Confirm Session Context

Present the agent's findings (decisions, blockers, notes) and end with an explicit question:

"Does this look right? Reply yes to continue, or tell me what to add or change."

If a category is empty, omit it.

STOP. Wait for explicit confirmation before proceeding to Step 7. If the user mentions new bugs or issues during this step, add them to the checkpoint notes — do NOT investigate or fix them.

Step 7: Generate Continuation Prompt

Rules:

  • Always include <context>, <current_state>, <next_action>, <key_files>. Omit <decisions>, <blockers>, <notes> if empty.
  • No absolute paths with usernames → use relative paths. No secrets/credentials → use placeholders.

Create a continuation prompt following the template in checkpoint-template.md.

Step 8: Save Checkpoint

Check if $PROJECT_ROOT/.dev/$FEATURE_NAME/checkpoint.md already exists. Remember whether the file existed as $IS_FIRST_CHECKPOINT (true if the file did NOT exist, false if it did).

If it exists, read it first (the Write tool requires reading before overwriting). Then write the continuation prompt to that path.

Step 9: Summary

Report:

  • Which feature was checkpointed
  • PRD updates made (list each file and what was changed, or state "No updates needed")
  • What the next steps are
  • Confirm the checkpoint location

Step 10: Workflow Setup (First Checkpoint Only)

REQUIRED CHECK: Evaluate $IS_FIRST_CHECKPOINT (set in Step 8). - If $IS_FIRST_CHECKPOINT is true (this is a NEW checkpoint — no checkpoint.md existed before Step 8): execute this step. - If $IS_FIRST_CHECKPOINT is false (checkpoint.md already existed): skip to Step 11.

This step offers the user a worktree or branch for the feature. Follow the instructions in worktree-guide.md.

Use $WORKTREE as the absolute path to scripts/worktree-setup.sh within this skill's directory. Apply the path safety rules from Step 0.

Step 11: Optional Commit

Skip this step entirely if ANY of these are true:

  • This is not a git repository
  • git status --porcelain output is empty (no uncommitted changes)

Note: Run git status --porcelain fresh here — do NOT reuse Step 5's result, because Step 9.5 may have moved files.

If there are uncommitted changes, generate a commit message from the checkpoint context:

  • Format: <Summary of what was accomplished this session>
  • Derive the summary from the checkpoint's <context> and <current_state> sections
  • Keep it to one concise sentence (under 72 characters if possible)

STOP. Present the following to the user and wait for their response:

Ready to commit your changes: `` <output of git status --short> ` Proposed commit message: ` <generated commit message> `` Commit these changes?

If the user declines: End the skill normally — no further action.

If the user accepts, run:

git add -u
# If there are user-approved untracked files from `git status --short`, add them explicitly:
# git add -- "<path>"
git commit -m "<generated commit message>"

Confirm with git log -1 --oneline and report the commit hash.

Wrap-Up Suggestion

After the checkpoint is complete (whether or not a commit was made), suggest:

Before ending your session, consider running /dev-wrapup to review learnings worth persisting and identify self-improvement signals.

平台分布

Codex

36.81%
按下载量换算41

Claude

30.77%
按下载量换算34

Cursor

16.65%
按下载量换算18

Gemini CLI

8.37%
按下载量换算9

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

执行命令

安装流程涉及命令执行,可能通过 npx skills add https://github.com/andreaserradev-gbj/dev-workflow --skill dev-checkpoint 联网下载 Skill 或依赖。用户安装前应确认命令来源、仓库内容和执行环境。

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