Analyzes update_report.md and report_structure.md to recommend relevant MCP servers and Claude skills. Recommends two categories - data collection tools (domain-specific databases, APIs, search tools) and report enhancement tools (visualization, analysis, formatting). Use when the user wants tool or skill recommendations based on their research report. Triggers include requests to "recommend skills", "suggest tools for my research", "what MCP servers should I use", or "find useful skills for this research".
sa-akinori0 estrellas21 mar 2026
Ocupación
Categorías
Depuración
Contenido de la habilidad
Overview
This skill analyzes the content of update_report.md and report_structure.md to recommend relevant MCP servers and Claude skills that would support the research workflow. It categorizes recommendations into:
Category A (Data Collection): Tools for gathering research data (academic databases, APIs, search tools, data repositories)
Category B (Report Enhancement): Tools for visualization and report quality (plotting libraries, diagram tools, analysis frameworks)
It uses web search to find appropriate tools and presents numbered recommendations to the user.
Workflow
Follow these steps in order:
Step 1: Check for input files
Check for both update_report.md and report_structure.md in the current working directory using the Read tool.
Priority order:
If both files exist: Read both files (recommended for best recommendations)
Skills relacionados
If only report_structure.md exists: Read it and proceed (can identify specific data/visualization needs)
If only update_report.md exists: Read it and proceed (can identify general research domain)
If neither file exists:
Inform the user that neither file was found
Suggest using research-report-enhancer and research-report-structure-planner first
STOP the workflow here
Read the available file(s) using the Read tool and proceed to Step 2.
Step 2: Ask user about recommendation preferences
Use the AskUserQuestion tool to ask the user about preferences for each category:
Question 1: Category A - Data Collection Tools
How many data collection tools (academic databases, APIs, search engines, data repositories, etc.) do you need?
Options:
2-3 recommendations
4-5 recommendations
6-8 recommendations
10+ recommendations
Question 2: Category B - Report Enhancement Tools
How many report enhancement tools (plotly, matplotlib, data visualization, diagram generation, etc.) do you need?
Options:
2-3 recommendations
4-5 recommendations
6-8 recommendations
10+ recommendations
None (I only need data collection tools)
Question 3: Tool Type Preference
What type of tools should be recommended?
Options:
Both MCP servers and Claude skills (recommended)
Claude skills only
MCP servers only
Step 3: Analyze input files content
Based on the available file(s), identify different aspects:
From update_report.md (if available):
Research domain (e.g., machine learning, financial analysis, social sciences, engineering, linguistics, environmental science)
Research topics and questions (identify key themes, methodologies, research questions)
General data needs (e.g., academic papers, datasets, APIs, public records, survey data)
Never assume specific tools - always search based on identified domain and needs
Use search queries like: "[identified need] MCP server 2026" or "Claude Code [identified need] skill"
Look for actively maintained tools with good documentation
Prioritize tools that match the specific needs identified in Step 3
Let the research domain and requirements guide the search, not pre-defined categories
Step 5: Generate categorized recommendations
Create a numbered list of recommendations organized by category:
========================================
Recommended Tools for Your Research
========================================
CATEGORY A: DATA COLLECTION TOOLS
------------------------------------------
Claude Skills:
1. [Skill Name]
Description: [What it does]
Why useful: [How it helps collect/access research data]
URL: [Installation URL or search guidance]
MCP Servers:
2. [MCP Server Name]
Description: [What data it provides]
Why useful: [How it relates to the research data needs]
Installation: [npm/git/python and package details]
[Continue for Category A based on user's requested count...]
CATEGORY B: REPORT ENHANCEMENT TOOLS
------------------------------------------
Claude Skills:
[N]. [Skill Name]
Description: [What visualization/analysis it provides]
Why useful: [How it improves report quality - reference specific figures from report_structure.md]
URL: [Installation URL or search guidance]
MCP Servers:
[N+1]. [MCP Server Name]
Description: [What visualization/processing capability]
Why useful: [How it addresses specific report needs]
Installation: [npm/git/python and package details]
[Continue for Category B based on user's requested count...]
Important formatting:
Clearly separate Category A and Category B
Within each category, separate Claude skills and MCP servers
Number all items sequentially starting from 1
For Category B, explicitly reference figure types from report_structure.md when applicable
Include URLs or installation instructions when available
If URLs aren't found, provide clear search guidance
Step 6: Present recommendations and ask for user input
Display all recommendations clearly to the user
Ask the user which items they would like to include in an install-skills.txt file
✓ Each line has exactly 3 space-separated parts (type, name/URL, URL/type)
✓ GitHub URLs use mcp <URL> git format (NOT skill format)
✓ Only .skill download URLs use skill <name> <URL> format
✓ npm packages use mcp <package> npm format
✓ Python packages use mcp <package> python format
✓ No lines with just skill/package names without URLs/types
✓ Comments start with # and are ignored by the parser
Common mistakes to AVOID:
❌ DON'T write: some-skill-name (missing format entirely)
❌ DON'T write: skill repo-name https://github.com/org/repo.git (GitHub repos are not .skill files)
✓ DO write: mcp https://github.com/org/repo-name.git git
❌ DON'T write: skill some-tool (missing URL and format)
✓ DO write: mcp https://github.com/org/some-tool.git git (for repos)
✓ DO write: skill some-tool https://example.com/some-tool.skill (for .skill files)
After creating the file:
Use Write tool to create mcp-servers/install-skills.txt
Inform the user that:
mcp-servers/install-skills.txt has been created with the CORRECT format for skill-mcp-installer
They should use skill-mcp-installer skill to batch install all items (DO NOT manually install)
skill-mcp-installer will automatically:
Clone git repositories to appropriate directories
Install npm packages locally
Install Python packages locally
Generate .claude/mcp_config.json with correct entry points
All installations will be local to the project directory (./skills/ and ./mcp-servers/)
IMPORTANT: Always double-check the format before writing. The skill-mcp-installer expects this exact format and will fail with cryptic errors if the format is incorrect.
Important Notes
Always check for both files: Read both update_report.md and report_structure.md when available for best recommendations
Never assume file contents: Always use Read tool to check file contents
Category A (Data Collection): Focus on tools that provide access to research data sources
Category B (Report Enhancement): Focus on visualization, analysis, and report quality tools
When report_structure.md is available, reference specific figure types (e.g., "戦略マップ", "比較チャート") in recommendations
Use WebSearch to find current, relevant tools (search results should be from 2026 or recent)
Prioritize tools that directly address the research needs identified in the files
For MCP servers, provide clear installation/configuration instructions
Both Claude skills and MCP servers can be included in install-skills.txt using the updated format