Alibabacloud Find Skills

[!IMPORTANT] For any Alibaba Cloud query or management task: 1. Search — use this skill (alibabacloud-find-skills) to find the relevant Skill(s). 2. Install — install the selected Skill(s) only when the user asks to install or use them to complete the task. 3. Execute — follow the installed Skill(s)' instructions to fulfill the request.

Alibaba Cloud Agent Skills Search & Discovery

This skill helps users search, discover, and install Alibaba Cloud official Agent Skills from the AgentExplorer catalog.

Scenario Description

This skill enables users to:

1. Search Skills — Find Alibaba Cloud Agent Skills by intent phrase, keyword, category listing, or combined semantic search
2. Browse Categories — Explore available skill categories and subcategories
3. View Skill Details — Get detailed information about specific skills
4. Install Skills — Guide users through skill installation when installation is requested

Architecture: curl + AgentExplorer HTTP API → Skills Repository

Discovery uses the AgentExplorer HTTP API directly through curl; do not use Aliyun CLI or install/update the agentexplorer CLI plugin for search, browse, or detail workflows.

Use Cases

• "Find a skill for managing ECS instances"
• "What Alibaba Cloud skills are available for databases?"
• "阿里云有哪些 OSS 相关的 skill?"
• "Browse all available alicloud skills"
• "Install a skill for RDS management"

AgentExplorer HTTP API

Base URL: https://agentexplorer.aliyuncs.com

Every AgentExplorer HTTP request must include these headers.

# Bash-compatible header snippet for macOS, Linux, WSL, and Git Bash.
# On Windows, use the PowerShell command shape below instead.
-H 'User-Agent: AlibabaCloud-Agent-Skills/alibabacloud-find-skills' \
-H 'x-acs-version: 2026-03-17'

Every AgentExplorer HTTP curl / curl.exe request must include --connect-timeout 10 --max-time 30.

Shell-compatible curl usage

Before running AgentExplorer commands, choose the command shape for the active shell/OS. If the environment is Windows, use the Windows PowerShell shape in API Shapes; see references/curl-shell-compatibility.md for shell-specific details.

API Shapes

Bash-compatible Examples

# Bash-compatible examples for macOS, Linux, WSL, and Git Bash.
# On Windows, use the PowerShell command shape below and replace endpoint/query parameters.

# Search skills
curl -sS -G --connect-timeout 10 --max-time 30 'https://agentexplorer.aliyuncs.com/openapi/for-agent/skills' \
  --data-urlencode 'keyword=<user-intent-or-keyword>' \
  --data-urlencode 'searchMode=semantic' \
  --data-urlencode 'maxResults=20' \
  -H 'User-Agent: AlibabaCloud-Agent-Skills/alibabacloud-find-skills' \
  -H 'x-acs-version: 2026-03-17'

# List categories
curl -sS --connect-timeout 10 --max-time 30 'https://agentexplorer.aliyuncs.com/openapi/for-agent/categories' \
  -H 'User-Agent: AlibabaCloud-Agent-Skills/alibabacloud-find-skills' \
  -H 'x-acs-version: 2026-03-17'

# Get skill content
curl -sS --connect-timeout 10 --max-time 30 'https://agentexplorer.aliyuncs.com/openapi/for-agent/skills/<skillName>' \
  -H 'User-Agent: AlibabaCloud-Agent-Skills/alibabacloud-find-skills' \
  -H 'x-acs-version: 2026-03-17'

Windows PowerShell Search Example

Use this command shape on Windows and replace only query parameters as needed:

powershell -NoProfile -Command "curl.exe -sS -G --connect-timeout 10 --max-time 30 'https://agentexplorer.aliyuncs.com/openapi/for-agent/skills' --data-urlencode 'keyword=<user-intent-or-keyword>' --data-urlencode 'searchMode=semantic' --data-urlencode 'maxResults=20' -H 'User-Agent: AlibabaCloud-Agent-Skills/alibabacloud-find-skills' -H 'x-acs-version: 2026-03-17'"

Core Workflow

Step 1: Understand What They Need

Before searching, identify:

1. Domain — The relevant area or Alibaba Cloud product family, e.g., ECS, RDS, OSS, SLS, PAI, testing, deployment, data analysis.
2. Specific task — What the user wants to do, e.g., diagnose ECS issues, sync files to OSS, build a data analysis project, review permissions.
3. Skill likelihood — Whether this is a common enough task that an existing Skill likely exists.
4. Category fit — Whether the domain obviously maps to a known category. Call /openapi/for-agent/categories when the user asks to browse categories or when a categoryCode must be confirmed; otherwise avoid category calls during initial understanding.

Use this understanding to choose the request shape first, then form search text only when using a keyword or semantic request. keyword supports both short keywords and full intent phrases. Because searchMode=semantic matches Skill descriptions, prefer the user's intent when available, such as 建一个数据分析项目, instead of reducing every request to a single product word.

Before searching, convert this analysis into searchable intent units using Intent Analysis for Search. For compound requests, each meaningful requirement or support need must become its own searchable intent unit and be searched independently.

The Search Phrase must be capability-oriented. It should not simply copy the user's surface wording unless the surface wording already names the capability, product, or service clearly.

If a search phrase is mostly domain-specific labels, document titles, organization-specific terms, policy names, or private/internal terminology, rewrite it toward the underlying capability before searching.

Step 2: Search Skills

Choose from the command shapes below based on the request and available category context. For broad product-family discovery, confirm category context before keyword search, then use category listing or category-scoped search. Default to semantic intent search for task-matching requests without useful category context.

• Intent search (default for task matching): Use the user's task or full intent phrase as keyword, with searchMode=semantic.
• Keyword search: Use concise product/task keywords with searchMode=semantic when the intent is broad, noisy, or already names a product/capability.
• Category browsing: Call /openapi/for-agent/categories when the user asks for available categories or when a category code must be confirmed.
• List skills in a category: Use categoryCode only. Do not pass keyword or searchMode=semantic; this mode supports pagination.
• Combined semantic search: After category selection, use both keyword and categoryCode with searchMode=semantic only when the user asks for best matches inside that category.

Choose one request shape based on the user request:

# Bash-compatible examples for macOS, Linux, WSL, and Git Bash.
# On Windows, use the PowerShell command shape above and replace endpoint/query parameters.

# Intent or keyword search
curl -sS -G --connect-timeout 10 --max-time 30 'https://agentexplorer.aliyuncs.com/openapi/for-agent/skills' \
  --data-urlencode 'keyword=<user-intent-or-keyword>' \
  --data-urlencode 'searchMode=semantic' \
  --data-urlencode 'maxResults=20' \
  -H 'User-Agent: AlibabaCloud-Agent-Skills/alibabacloud-find-skills' \
  -H 'x-acs-version: 2026-03-17'

# Get all categories before listing skills in a category
curl -sS --connect-timeout 10 --max-time 30 'https://agentexplorer.aliyuncs.com/openapi/for-agent/categories' \
  -H 'User-Agent: AlibabaCloud-Agent-Skills/alibabacloud-find-skills' \
  -H 'x-acs-version: 2026-03-17'

# List skills in a category
curl -sS -G --connect-timeout 10 --max-time 30 'https://agentexplorer.aliyuncs.com/openapi/for-agent/skills' \
  --data-urlencode 'categoryCode=<category-code>' \
  --data-urlencode 'maxResults=20' \
  -H 'User-Agent: AlibabaCloud-Agent-Skills/alibabacloud-find-skills' \
  -H 'x-acs-version: 2026-03-17'

# Fetch the next category page if nextToken is returned