Deep research

Skill FMATheNomad/deep-research-skill/skills/deep-research

Massive autonomous web research for AI coding agents. ⭐ Free Firecrawl alternative. No API keys. Open source. Sponsor ❤️

Install

npx -y skills add FMATheNomad/deep-research-skill --skill deep-research

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Massive autonomous web research via AI-powered browsing. Use this skill when the user needs to research a topic deeply, compare information across multiple sources, gather data from many websites, or conduct comprehensive web research. Automatically searches, fetches dozens of pages in parallel, and synthesizes findings. Triggers: 'research this topic', 'search for complete info about', 'deep research', 'compare X and Y', 'gather data from multiple sources', 'research about'. Far more efficient than browser automation for research — uses direct content fetching in parallel.

SKILL.md

5.6 KB, as published. Nobody here has run it

Deep Research — Autonomous Massive Web Research

A pipeline for conducting deep, multi-source web research. Uses OpenCode's built-in websearch and webfetch tools to search, fetch dozens of pages in parallel, and synthesize comprehensive findings.

No API keys needed. No accounts. Completely free.

How It Works

User: "Search for complete info about antimatter"

Agent:
  Step 1: websearch("antimatter physics")          → 10+ source URLs
  Step 2: wefetch(source1) + webfetch(source2) + ... (parallel, 10+ at once)
  Step 3: Read & synthesize all content
  Step 4: Present structured summary with sources

Research Workflow

Step 1: Understand the Scope

Determine:

Topic — what to research
Depth — overview vs deep dive vs comparison
Sources needed — general web, academic, news, documentation
Output format — summary, comparison table, structured report

Step 2: Search for Sources

Use websearch to find relevant pages. Always search multiple queries for comprehensive coverage:

# Primary search — cast a wide net
websearch("topic")
websearch("topic research paper")
websearch("topic latest 2025 2026")
websearch("topic comparison alternatives")
websearch("topic tutorial guide")

# Collect ALL returned URLs across searches

Pro tip: Use OPENCODE_ENABLE_EXA=1 environment variable before starting OpenCode to enable websearch tool.

Step 3: Fetch Pages in Parallel

Use webfetch to fetch all collected URLs. Fetch at least 10-20 pages in parallel for proper deep research:

# Launch all fetches in parallel — do NOT wait for each one
webfetch(url1)
webfetch(url2)
webfetch(url3)
...
webfetch(urlN)

Rules:

Fetch ALL URLs, not just a few
Prioritize: Wikipedia, academic sources, official docs, reputable blogs
Skip obvious low-value pages (forums, spam)
If some fail, note it and continue

Step 4: Analyze & Synthesize

After receiving all content:

Identify key themes across sources
Note contradictions — different sources may disagree
Find unique insights — each source adds value
Structure the answer:
- Executive summary (2-3 sentences)
- Detailed findings with source citations
- Key statistics / data points
- Different perspectives
- Conclusion

Step 5: Present Results

Format the response with:

## 📋 Research: [Topic]

**Summary:** [2-3 sentence overview]

### Key Findings
1. **[Finding 1]** — [detail] ([source](url))
2. **[Finding 2]** — [detail] ([source](url))
...

### Detailed Analysis
[Comprehensive synthesis of all sources]

### Sources
- [Title](url) — [what this source contributed]
- [Title](url) — [what this source contributed]
...

### Further Questions
- [Aspect not fully covered]
- [Related topic worth exploring]

Research Templates

Template: Comparison Research

# User: "Compare X and Y"
websearch("X vs Y comparison")
websearch("X features pricing")
websearch("Y features pricing")
websearch("X review")
websearch("Y review")

# Fetch all results in parallel
webfetch(url1) ... webfetch(urlN)

# Present as comparison table:
# | Feature | X | Y |
# | Pricing | $ | $ |
# | Pros | ... | ... |
# | Cons | ... | ... |

Template: Deep Technical Research

# User: "How does technology X work?"
websearch("X explained")
websearch("X architecture")
websearch("X tutorial")
websearch("X whitepaper")
websearch("X vs alternatives")

Template: Market / Competitive Research

# User: "Analyze the market for X"
websearch("X market size 2025")
websearch("X competitors")
websearch("X trends")
websearch("X funding")
websearch("X review")

Tips for Maximum Results

Search multiple queries — different angles yield different sources
Fetch everything in parallel — bash & style or sequential tool calls
Prioritize quality sources — Wikipedia, official docs, .edu, .gov, reputable tech sites
Note failed fetches — try alternative URLs if available
Cite sources — always link back to original content
10+ pages minimum — for truly deep research, aim for 20-30+ pages

How This Beats Firecrawl

Feature	Firecrawl	This Skill
API Key Required	✅ Yes	❌ No
Cost	Free tier (500K credits)	Free unlimited
Search	✅ Built-in	✅ via `websearch` (Exa)
Markdown output	✅	✅ via `webfetch`
Parallel fetching	✅	✅ (agent can batch)
AI synthesis	Separate API	✅ Built into agent
Customizable	Limited	✅ Full control
Open source	AGPL	✅ MIT

Limitations

websearch requires Exa (enabled via OPENCODE_ENABLE_EXA=1 or OpenCode provider)
webfetch fetches one URL at a time per tool call (but agent can parallelize)
Rate limits depend on your OpenCode provider
Some sites block automated fetching (JavaScript-heavy SPAs)