Token导航 LogoToken导航

Agent Skill · 研究检索

build-session-context(构建会话上下文)

build-session-context 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。

build-session-context 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,权限为只读;许可证为 MIT,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 69,749。

总安装 216周安装 9GitHub Stars 1下载 72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
MIT

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:build-session-context(构建会话上下文)
来源仓库:https://github.com/dudusoar/vrp-toolkit
仓库路径:skills/build-session-context
安装命令:
npx skills add https://github.com/dudusoar/vrp-toolkit --skill build-session-context
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。不同来源提供的安装方式可能略有差异;本站展示可直接复制的安装命令,安装前请核对来源页面。

skills.shnpx skills
npx skills add https://github.com/dudusoar/vrp-toolkit --skill build-session-context

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。

build-session-context 从项目日志中提取关键信息,生成简洁上下文摘要以加速会话恢复。适用于多文件协作项目,帮助 Agent 快速定位当前阶段与待办事项。优先读取 CLAUDE.md 与进度文件,提取模块状态与更新时间等核心元数据。需确保项目存在标准化文档结构,避免因文件格式混乱导致信息遗漏。

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/build-session-context

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/dudusoar/vrp-toolkit --skill build-session-con…;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

    它标注为只读,先用只读任务试一次,确认行为符合预期再扩大使用范围。

常见问题

6 条

build-session-context 怎么安装?

在终端执行 npx skills add https://github.com/dudusoar/vrp-toolkit --skill build-session-context。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

build-session-context 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。

build-session-context 需要什么权限?

权限级别是只读。该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

build-session-context 是免费的吗?

来源标注的许可证是 MIT。本站不代收任何费用,也不托管安装包;能否商用、要不要署名,以许可证原文为准。

build-session-context 用的人多吗?

公开数据是累计安装 216 次,近一周安装 9 次,下载 72 次,GitHub 1 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。

build-session-context 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。
  • 3 个目录同时收录,说明这个 Skill 流传比较广。
  • 许可证明确为 MIT,商用和分发边界好判断。
  • 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

  • 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。
  • 需要许可证清楚、并且有第三方扫描结果的团队项目。

不太适合

  • 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
build-session-context

构建会话上下文 · 本页

2161只读MIT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

Build Session Context

Quickly resume work by extracting key information from project logs and presenting a concise context summary for efficient session startup.

Goal

Extract essential information from multiple project documentation files and present a concise, token-efficient summary that enables quick context recovery without reading entire files.

Workflow

Step 1: Extract Key Information from Each File

Read each file and extract only the most critical information:

1. CLAUDE.md (Project Entry Point)

  • Phase: Current project phase (e.g., "Phase 2 - Refactoring")
  • Last Updated: Date of last update
  • Project Status: High-level overview (e.g., "9/9 files migrated, 90% complete")
  • Note: CLAUDE.md is now a clean entry point; detailed tasks are in TASK_BOARD.md

2. TASK_BOARD.md (Detailed Task Tracking)

  • Completed Tasks: Count and 2-3 most recent
  • In Progress Tasks: List (limit to 3 most important)
  • Next Steps: Top 2-3 priorities
  • Blockers: Any active blockers (critical)
  • Progress Metrics: Overall percentages if available

3. MIGRATION_LOG.md (Migration History)

  • Progress: X/9 files completed (percentage)
  • Recent Migrations: Last 2-3 entries (date, title, status)
  • Index: Check index for quick overview if available

4. DEBUG_LOG.md (Problem Tracking)

  • Active Issues: Count and brief descriptions
  • Recent Resolutions: Last 2-3 resolved issues
  • Urgent Blockers: Any issues marked as critical

5. GIT_LOG.md (Change History)

  • Recent Commits: Last 3-5 commits (hash, message)
  • Activity Pattern: Type of recent work (features, fixes, docs)
  • Last Commit Date: When was last work done

6. SKILLS_LOG.md (Skills Changes) [Optional]

  • Recent Changes: Last 1-2 skill updates (if within last 7 days)
  • New Skills: Any skills added recently
  • Only include if skills changed recently to keep context minimal

7. Git Status (Current Workspace)

  • Current Branch: Which branch is active
  • Uncommitted Changes: Any staged/unstaged changes
  • Untracked Files: New files not yet in git

Step 2: Generate Concise Context Summary

Format the extracted information into a compact summary:

## 🚀 Session Start - VRP Toolkit
**Phase:** [Phase] | **Last Updated:** [Date] | **Branch:** [branch]

## 📊 Status Snapshot
- **Migration:** X/9 files (XX%) | **Active Issues:** [count]
- **Completed:** [count] tasks | **In Progress:** [count] tasks
- **Blockers:** [None/list]

## 🔍 Recent Activity
**Git:** [Last 3 commits, one line each]
**Migrations:** [Last 2 migration titles]
**Issues:** [Active issue count, recent resolution]

## 🎯 Current Focus
1. [Primary in-progress task]
2. [Secondary in-progress task]

## 📋 Immediate Next Steps
1. [Highest priority next action]
2. [Secondary next action]

## ⚠️ Active Blockers (if any)
- [Blocker 1]
- [Blocker 2]

Token Optimization Guidelines:

  • Keep total summary under 1000 tokens if possible
  • Use abbreviations: e.g., "ALNS" not "Adaptive Large Neighborhood Search"
  • Limit lists to 3-5 items maximum
  • Use bullet points over paragraphs
  • Prioritize recent information over historical

Step 3: Suggest Context-Aware Next Action

Based on the summary, recommend the most appropriate next step:

Decision Matrix:

  • Blockers present → Address most critical blocker first
  • Uncommitted changes → Review and commit before new work
  • Active issues in DEBUG_LOG.md → Continue debugging if in progress
  • Migration in progress → Continue with current migration
  • No active tasks → Start highest priority from "Next Steps"
  • Long break (>3 days) → Review recent commits first

Be specific and actionable:

  • Instead of: "Work on migration"
  • Use: "Continue migrating operators.py to vrp_toolkit/algorithms/alns/operators.py"

Step 4: Offer Skill Integration

After presenting summary, suggest relevant skills:

Common integrations:

  • update-task-board - Update task status if work completed
  • git-log - Commit uncommitted changes
  • log-debug-issue - Document issues encountered
  • update-migration-log - Log completed work
  • migrate-module - Continue file migration

Special Cases & Optimization

Long Log Files

When logs are very long (e.g., MIGRATION_LOG.md > 500 lines):

  • Rely on index section if available
  • Read only first/last few entries
  • Use grep or search for recent dates
  • Extract statistics rather than details

Missing Log Files

If a log file doesn't exist:

  • Note its absence in summary
  • Suggest creating it if needed
  • Proceed with available information

First Session After Break

If last update > 3 days ago:

  • Emphasize git history to rebuild context
  • Highlight any stale in-progress tasks
  • Suggest verifying environment/dependencies

High Token Count Situation

If summary is too long:

  • Further trim lists (2 items instead of 3)
  • Remove less critical sections
  • Use more abbreviations
  • Focus only on current/next actions

Integration with Project Documentation

TASK_BOARD.md: Detailed task status managed by update-task-board skill. This is the primary source for task tracking.

MIGRATION_LOG.md: Migration history maintained by update-migration-log skill. Provides evidence for task completion.

DEBUG_LOG.md: Issue tracking maintained by log-debug-issue skill. Source for active blockers.

GIT_LOG.md: Commit history maintained by git-log skill. Shows recent development activity.

SKILLS_LOG.md: Skills changes maintained by manage-skills skill. Optional context for recent workflow updates.

CLAUDE.md: High-level project overview (now simplified, points to other files for details).

Documentation Structure (post-refactoring):

  • CLAUDE.md → Entry point (~308 lines reduced)
  • TASK_BOARD.md → Task tracking (extracted from CLAUDE.md)
  • SKILLS.md → Skills reference (extracted from CLAUDE.md)
  • MIGRATION_GUIDE.md → Migration details (in migrate-module skill)

Example Output

## 🚀 Session Start - VRP Toolkit
**Phase:** Phase 2 - Refactoring (90% complete) | **Last Updated:** 2026-01-03 | **Branch:** main

## 📊 Status Snapshot
- **Migration:** 9/9 files (100%) | **Active Issues:** 0
- **Tasks:** 20+ completed | 1 in progress
- **Blockers:** None
- **Skills:** 10 total (recently refined documentation structure)

## 🔍 Recent Activity
**Git:** 4569fa1 feat(architecture): unified Solver interface
        9590014 feat(setup): make package installable
        0287f90 docs: comprehensive README
**Tasks:** Test suite creation (in progress)
**Docs:** CLAUDE.md simplified (~308 lines → dedicated files)
          Created TASK_BOARD.md, SKILLS.md, MIGRATION_GUIDE.md
**Skills:** manage-skills created, manage-python-env fixed compliance

## 🎯 Current Focus
1. Creating comprehensive test suite for new architecture

## 📋 Immediate Next Steps
1. Complete test suite with edge cases and integration tests
2. Test ALNSSolver with other VRPProblem implementations

## ⚠️ Active Blockers
None - ready to proceed

---
**💡 Suggested Action:** Continue test suite development or start with `update-task-board` to sync latest progress

Reference

For detailed file structures:

平台分布

Claude Code

28.78%
按下载量换算21

windsurf

21.04%
按下载量换算15

trae

16.77%
按下载量换算12

OpenCode

14.09%
按下载量换算10

Codex

8.36%
按下载量换算6

github-copilot

3.55%
按下载量换算3

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

只读

该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

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。

继续浏览同类 Skills