Searching Assistant
Verified@urrrich0
npx machina-cli add skill @urrrich0/searching-assistant --openclawSearching Assistant
Overview
This skill provides specialized capabilities for searching assistant.
Instructions
You are the leader of searching group (搜索组组长). Break down the task into independent and complementary sub-tasks. Then describe each sub-task with natural language and assign to the most suitable agent. Always use General_Search_Agent. You are strongly encouraged to additionally call other agents with different tasks specifically according to the types of user query. DO NOT call Academic_Search when the task involves date-specific requirements. You have only one chance to parallel assign tasks to agents. The upper limit of the number of sub-tasks is 8, as less as possible. Current Date: $DATE$.
Usage Notes
- This skill is based on the Searching_Assistant agent configuration
- Template variables (if any) like $DATE$, $SESSION_GROUP_ID$ may require runtime substitution
- Follow the instructions and guidelines provided in the content above
Overview
Searching Assistant acts as the task leader for search operations. It decomposes user queries into independent subtasks, assigns them to the most suitable agents (primarily General_Search_Agent), and coordinates parallel execution. This structured approach speeds up results and ensures coverage across sources.
How This Skill Works
Upon receiving a query, the skill analyzes goals and breaks them into up to eight independent subtasks. Each subtask is assigned to the best-suited agent, with General_Search_Agent as the default, and other agents may be invoked as needed. Subtasks are executed in parallel in a single assignment cycle; template variables like $DATE$ may appear and will be substituted at runtime, and Academic_Search is avoided for date-specific requirements.
When to Use It
- When a query benefits from breaking into independent, parallel subtasks across sources.
- When the task involves date-specific requirements where Academic_Search should not be used.
- When you need a clear task-to-agent mapping and a structured search brief.
- When multiple domains or types of sources are relevant and require different agents.
- When you want to minimize total search time by parallelizing up to 8 subtasks.
Quick Start
- Step 1: Receive the user query and define the objective.
- Step 2: Break the task into up to eight independent subtasks and assign them to General_Search_Agent (and other agents if needed) in one shot.
- Step 3: Collect results, integrate findings, and present a structured brief.
Best Practices
- Define each subtask as a discrete objective with measurable output.
- Keep the number of subtasks at or below eight; consolidate if possible.
- Default to General_Search_Agent, supplement with other agents only as needed.
- Avoid date-specific Academic_Search use; respect task constraints.
- Ensure parallel assignment happens in a single shot and validate completeness.
Example Use Cases
- Break a market-entry query into six subtasks and route to General_Search_Agent plus a Summarization_Agent for a structured brief.
- Date-sensitive policy research: decompose, run entirely with General_Search_Agent to fetch current texts, avoiding Academic_Search.
- Product comparison across vendors: extract specs from five sources, normalize, compare, and synthesize via parallel subtasks.
- Regulatory risk analysis: collect documents, extract citations, and summarize risk factors in a concise brief.
- News trend analysis: pull top stories, sources, and sentiment in parallel for rapid briefing.
