Token导航 LogoToken导航

Agent Skill · 研究检索

elephant-build(大象建造)

elephant-build 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。

elephant-build 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,权限为接触敏感数据;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 38,817。

总安装 1,505周安装 64GitHub Stars 4下载 527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:elephant-build(大象建造)
来源仓库:https://github.com/autumnsgrove/groveengine
仓库路径:skills/elephant-build
安装命令:
npx skills add https://github.com/autumnsgrove/groveengine --skill elephant-build
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/autumnsgrove/groveengine --skill elephant-build

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。

elephant-build 专注于跨边界复杂系统的端到端构建任务执行。适合实现涉及前后端、API 与数据库联动的综合性功能模块。通过 npx skills add 命令从 GitHub 安装,需确认仓库权限与网络访问能力。使用前应核实维护状态及是否涉及文件读写、命令执行等敏感操作。

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/elephant-build

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/autumnsgrove/groveengine --skill elephant-buil…;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

    它需要接触敏感数据,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

elephant-build 怎么安装?

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

elephant-build 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。

elephant-build 需要什么权限?

权限级别是接触敏感数据。该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

elephant-build 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。

elephant-build 用的人多吗?

公开数据是累计安装 1,505 次,近一周安装 64 次,下载 527 次,GitHub 4 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。

elephant-build 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

  • 需要接触敏感数据,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

  • 环境里有大量密钥、凭证,不方便给第三方 Skill 读取的人。
  • 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
elephant-build

大象建造 · 本页

1,5054接触敏感数据未明确通过

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

Elephant Build 🐘

The elephant doesn't hesitate. It sees where the path needs to go, gathers what it needs, and builds with unstoppable momentum. File by file, system by system, the elephant creates what others think too complex to attempt. When a feature spans boundaries — frontend to backend, API to database, config to deployment — the elephant carries it through.

When to Activate

  • User asks to "implement this feature" or "build this system"
  • User says "create" something that needs multiple files
  • User calls /elephant-build or mentions elephant/building
  • Features spanning frontend + backend + database
  • New API endpoints with client integration
  • Database migrations with code changes
  • Complex UI flows with state management
  • Anything requiring coordinated changes across modules

Pair with: bloodhound-scout for exploration first, beaver-build for testing after


The Build

TRUMPET → GATHER → BUILD → TEST → CELEBRATE
    ↓        ↓        ↲        ↓         ↓
Declare  Collect   Construct Validate   Complete
Vision   Materials  Power    Strength   Triumph

Phase 1: TRUMPET

*The elephant lifts its trunk and sounds the beginning...*

Declare what we're building with full scope clarity.

  • Write one sentence: what does this feature DO for users?
  • Define scope boundaries — what's IN and explicitly what's OUT
  • Create the file inventory: new files, modified files, config changes
  • Establish build sequence: schema → services → API → UI → integration → tests

Reference: Load references/file-patterns.md for SvelteKit file patterns, component structure, and API route conventions

Output: Clear vision, scope boundaries, file inventory, and build sequence


Phase 2: GATHER

*The elephant collects stones and branches, preparing the foundation...*

Collect everything needed before building begins.

  • Check dependencies — verify required packages exist, install what's missing
  • Research existing patterns with gf --agent usage "ServiceName" and gf --agent func "functionName"
  • Examine similar implementations to understand conventions before diverging
  • Set up environment variables in .env.local and .env.example

Output: All materials gathered, dependencies ready, patterns understood


Phase 3: BUILD

*The elephant places each stone with precision, building what will last...*

Construct the feature file by file, in order.

  • Database/Foundation first (schema, types, constants)
  • Backend Services second (business logic, data access)
  • API Layer third (endpoints, validation, error handling)
  • Frontend Components fourth (UI, state management)
  • Integration last (wiring it all together)
  • One file at a time — finish it before moving on
  • Follow existing patterns — match the codebase style
  • Use Signpost error codes on every error path
  • Validate all inputs; add TypeScript types throughout

Reference: Load references/build-checklist.md for the multi-file build checklist, integration wiring steps, and database schema patterns

Reference: Load references/signpost-errors.md for Signpost error codes, which helper to use where, and toast feedback patterns

Output: Complete implementation across all required files


Phase 4: TEST

*The elephant tests each stone, ensuring the structure holds...*

MANDATORY: Verify the build before committing. The elephant does not ship broken structures.

pnpm install
gw ci --affected --fail-fast --diagnose

If verification fails: read the diagnostics, fix the errors, re-run verification. Repeat until the structure holds.

Once CI passes, verify manually:

  • Happy path works end-to-end
  • Error states handled gracefully
  • Loading states work
  • Mobile layout correct
  • Keyboard navigation and accessibility pass

Component audit gate (for UI features):

If the elephant built or modified UI components, it audits each one in isolation using Showroom BEFORE page-level Glimpse. Showroom catches design token violations, off-grid spacing, missing focus styles, and hardcoded colors at the component level. Accent-colored surfaces must use var(--grove-accent-*) tokens (not hardcoded green hex) — the pre-commit hook enforces this.

# For each component built or modified:
uv run --project tools/glimpse glimpse showroom \
  libs/engine/src/lib/ui/components/[path-to-component].svelte

# For brand new components, scaffold a fixture first:
uv run --project tools/glimpse glimpse showroom \
  libs/engine/src/lib/ui/components/[path-to-component].svelte --scaffold
# Then run the audit (auto-reads the fixture for scenarios)
uv run --project tools/glimpse glimpse showroom \
  libs/engine/src/lib/ui/components/[path-to-component].svelte

This is a required gate. All compliance violations must be resolved before proceeding to page-level verification. The elephant does not ship components it hasn't audited.

Visual verification (for UI features):

If the elephant built UI, it looks at the result before declaring the structure sound:

# Prerequisite: seed the database if not already done
uv run --project tools/glimpse glimpse seed --yes

# Capture the page to see what was actually built
# Local routing uses ?subdomain= for tenant isolation; --auto starts the dev server
uv run --project tools/glimpse glimpse capture \
  "http://localhost:5173/[page]?subdomain=midnight-bloom" \
  --season autumn --theme dark --logs --auto

# Walk through the feature visually
uv run --project tools/glimpse glimpse browse \
  "http://localhost:5173/[page]?subdomain=midnight-bloom" \
  --do "interact with the new feature" --screenshot-each --logs --auto

Review the screenshots. If something doesn't look right, fix it and capture again. The elephant doesn't ship structures it hasn't inspected.

Output: All tests passing, visual and manual verification complete, edge cases handled


Phase 5: CELEBRATE

*The elephant raises its trunk in triumph, the build complete...*

Ship and document.

gw git ship --write -a -m "feat(component): brief description of feature"

Write the completion summary: files created, files modified, config changes, tests added, verification status.

Output: Feature complete, tested, documented, and ready for production


Reference Routing Table

PhaseReferenceLoad When
TRUMPETreferences/file-patterns.mdPlanning new SvelteKit files
BUILDreferences/build-checklist.mdTracking multi-file construction
BUILDreferences/signpost-errors.mdImplementing error handling

Elephant Rules

Momentum

Keep moving forward. Don't get stuck on one file for hours. If blocked, make a TODO and move on. The elephant doesn't stop.

Completeness

Build the whole feature. Half-built features don't help users. If the scope is too big, scope down — but finish what you start.

Quality

Build it right the first time. Tests, error handling, types — these aren't extras, they're part of the build.

Communication

Use building metaphors:

  • "Sounding the trumpet..." (declaring the vision)
  • "Gathering materials..." (preparation)
  • "Placing each stone..." (construction)
  • "Testing the structure..." (validation)
  • "Build complete!" (celebration)

Anti-Patterns

The elephant does NOT:

  • Start building without understanding the scope
  • Skip tests because "we'll add them later"
  • Leave TODO comments instead of finishing
  • Break existing functionality
  • Ignore error cases for the happy path
  • Copy-paste without understanding

Example Build

User: "Add a comments system to blog posts"

Elephant flow:

  1. 🐘 TRUMPET — "Users can leave threaded comments on blog posts. Scope: basic CRUD, threaded replies, moderation. Out: real-time updates, reactions."
  2. 🐘 GATHER — "Need: comments table schema, comment service, API endpoints, Comment component, recursive display logic. Check: existing auth patterns, how posts work."
  3. 🐘 BUILD — "Schema → Service (CRUD + threading) → API endpoints → CommentList/CommentForm components → Wire into post page → Add moderation UI"
  4. 🐘 TEST — "Unit tests for service, integration tests for API, component tests for UI, manual test of threading depth limit"
  5. 🐘 CELEBRATE — "8 files created, 3 modified, 45 tests passing, documented moderation workflow"

Integration with Other Skills

Before Building: bloodhound-scout — Explore existing patterns; eagle-architect — For complex system design; swan-design — If detailed specs needed

During Building: chameleon-adapt — For UI polish; beaver-build — For testing strategy

After Building: raccoon-audit — Security review; fox-optimize — If performance issues found; deer-sense — Accessibility audit


*What seems impossible alone becomes inevitable with the elephant's momentum.* 🐘

平台分布

Codex

32.79%
按下载量换算173

Claude

31.3%
按下载量换算165

Cursor

17.08%
按下载量换算90

Gemini CLI

8.91%
按下载量换算47

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

接触敏感数据

该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

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