agentleFS
Sign inStart free
LeadGrowGTMLeadGrowGTM/research-process-builder/.agents/skills/research-agent/SKILL.md

research-agent

Skill70 starsChanged 2 months agoMITView on GitHub

Standardized web research agent using FireCrawl (deep scrape) + SerperDev (search) + gpt-4o-mini (extraction). Validated against ground truth with annealable scoring loop. Trigger: any research task that needs web-sourced structured data extraction.

  • Reads credentials

17 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

2 rules. 0 are common, 1 is distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 11 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Run against all GT companies → get scores per category

    line 57
  2. Very specific:

    All read from C:\Users\mitch\Everything_CC\.env:

    line 89

“Common” is measured: kept by at least 6 in 10 of the 11 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Research Agent — FireCrawl + SerperDev + 4o-mini

Standardized, low-cost web research agent. Combines three tools into one deterministic pipeline:

ToolRoleCost
SerperDevGoogle search — find relevant URLs + snippets~$0.001/search
FireCrawlDeep page scrape — extract clean markdown from top URLs~$0.001/scrape
gpt-4o-miniStructured extraction — classify, extract fields, validate~$0.00015/1K tokens

When to Use

Any task that needs web-sourced structured data:

  • Company profile extraction (description, size, HQ, funding)
  • People discovery (founders, executives, decision makers)
  • Competitive intelligence (competitors, market position)
  • Signal detection (hiring, funding, product launches)
  • Any research process in processes/find-*/

Architecture

Input (domain/company)
  │
  ├─► SerperDev search → URLs + snippets
  │
  ├─► FireCrawl scrape → clean markdown (top 3 URLs)
  │
  ├─► gpt-4o-mini extraction → structured JSON
  │
  └─► GT validator → accuracy score vs ground-truth/*.json

Usage

# Research a single company
py scripts/agent_research.py --domain stripe.com --category company_profile

# Research with ground truth validation
py scripts/agent_research.py --domain stripe.com --category company_profile --gt

# Batch run against all GT companies
py scripts/agent_research.py --all-gt --category founders_ceo

# Anneal: score against GT and identify failure patterns
py scripts/agent_research.py --all-gt --anneal --rounds 5

Annealing Loop

  1. Run against all GT companies → get scores per category
  2. Identify categories/companies scoring below 0.90
  3. Examine failure patterns (missing fields, wrong extraction, bad URLs)
  4. Adjust Serper queries or extraction prompt
  5. Re-run → compare scores → iterate until 90%+ across all GT

Cost per annealing round: ~$0.03 (10 companies × 3 searches + 3 scrapes + 1 extraction each)

Output Format

Results written to output/agent-research/{company}/{category}.json:

{
  "company": "Stripe",
  "domain": "stripe.com",
  "category": "company_profile",
  "model": "gpt-4o-mini",
  "timestamp": "2026-05-13T19:00:00Z",
  "sources": ["https://stripe.com/about", "https://en.wikipedia.org/wiki/Stripe"],
  "extracted": {
    "description": "Financial infrastructure platform...",
    "employee_count": "8,500-10,000",
    "headquarters": "South San Francisco, CA",
    "founded_year": "2010"
  },
  "gt_score": 0.95
}

Env Vars Required

All read from C:\Users\mitch\Everything_CC\.env:

  • FIRECRAWL_API_KEY
  • SERPER_API_KEY
  • OPENAI_API_KEY

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free