llm-knowledge-bases

Attribution: Originally authored by Weizhena. Included with attribution for use in the Deep query workflow.

Research Skill - Preliminary Research

Trigger Method

/research <topic>

Execution Flow

Step 1: Generate Initial Framework Using Model's Internal Knowledge

Based on the topic, use the model's existing knowledge to generate:

• A list of main research objects/items in the field
• A suggested framework of research fields

Output {step1_output} and use AskUserQuestion to confirm:

• Does the items list need additions or removals?
• Does the field framework meet the requirements?

Step 2: Web Search Supplement

Use AskUserQuestion to inquire about the time range (e.g., last 6 months, 2024 to present, no limit).

Parameter Collection:

• {topic}: Research topic input by user
• {YYYY-MM-DD}: Current date
• {step1_output}: Complete output content generated in Step 1
• {time_range}: Time range specified by user

Hard Constraint: The following prompt must be strictly reproduced, only replacing variables in {xxx}, no rewriting of structure or wording allowed.

Launch 1 web-search-agent (background), Prompt Template:

prompt = f"""## Task
Research Topic: {topic}
Current Date: {YYYY-MM-DD}

Based on the following preliminary framework, supplement the latest items and recommended research fields.

## Existing Framework
{step1_output}

## Objectives
1. Verify if existing items miss important objects
2. Supplement items based on missing objects
3. Continue searching for {topic}-related items within {time_range} and supplement
4. Supplement new fields

## Output Requirements
Return structured results directly (do not write files):

### Supplemented Items
- item_name: Brief explanation (why should it be included)
...

### Recommended Additional Fields
- field_name: Field description (why this dimension is needed)
...

### Information Sources
- [Source 1](url1)
- [Source 2](url2)
"""

One-shot Example (assuming research on AI Coding development history):

## Task
Research Topic: AI Coding Development History
Current Date: 2025-12-30

Based on the following preliminary framework, supplement the latest items and recommended research fields.

## Existing Framework
### Items List
1. GitHub Copilot: Developed by Microsoft/GitHub, first mainstream AI programming assistant
2. Cursor: AI-first IDE, based on VSCode
...

### Field Framework
- Basic Info: name, release_date, company
- Technical Features: underlying_model, context_window
...

## Objectives
1. Verify if existing items miss important objects
2. Supplement items based on missing objects
3. Continue searching for AI Coding Development History-related items within 2024 to present and supplement
4. Supplement new fields

## Output Requirements
Return structured results directly (do not write files):

### Supplemented Items
- item_name: Brief explanation (why should it be included)
...

### Recommended Additional Fields
- field_name: Field description (why this dimension is needed)
...

### Information Sources
- [Source 1](url1)
- [Source 2](url2)

Step 3: Inquire About User's Existing Fields

Use AskUserQuestion to ask if the user has predefined field files, read and merge if available.

Step 4: Generate Outline (Separate Files)

Merge {step1_output}, {step2_output} and user's existing fields to generate two files:

outline.yaml (items + configuration):

• topic: Research topic
• items: List of research objects
• execution:
• batch_size: Number of parallel agents (requires AskUserQuestion confirmation)
• items_per_agent: Number of items per agent for research (requires AskUserQuestion confirmation)
• output_dir: Results output directory (default ./results)

fields.yaml (field definitions):

• Field categories and definitions
• Each field's name, description, detail_level
• detail_level hierarchy: Minimal → Brief → Detailed
• uncertain: List of uncertain fields (preserved fields, automatically filled in deep phase)

Step 5: Output and Confirmation

• Create directory: ./{topic_slug}/
• Save: outline.yaml and fields.yaml
• Display to user for confirmation

Output Path

{current_working_directory}/{topic_slug}/
  ├── outline.yaml    # items list + execution configuration
  └── fields.yaml     # field definitions

Follow-up Commands

• /research-add-items - Supplement items
• /research-add-fields - Supplement fields
• /research-deep - Start deep research