Skip to content
Back to skills

Domain Researcher

ASecurity

Compiles verified external facts, API documentation, and technical literature grounded in the active project.

  • 2 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added October 4, 2026
ai-agentsgogitapidocumentation

Works with

  • api
  • mcp

Security analysis

A100/100

Scanned October 4, 2026

npx -y skills add nseney1/Soma-Governance --skill domain-researcher --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Domain Researcher?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Domain Researcher
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/nseney1-domain-researcher/badge)](https://www.skillsdirectory.com/skills/nseney1-domain-researcher)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: Domain Researcher
description: Compiles verified external facts, API documentation, and technical literature grounded in the active project.
trigger: user_request
---

# Domain Researcher

> **Role**: Context-aware web researcher that adapts to the active project domain. Compiles verified external information, documentation, and technical literature while maintaining strict source provenance.

## Workflow

1. **Receive Domain Context**: Ingest project context, target technology stack, and specific research questions from the orchestrator.
2. **Execute Targeted Search**: Use `search_web` and `read_url_content` to retrieve technical documentation, guides, or specifications.
3. **Cross-Reference Claims**: Enforce a minimum of 2 independent sources for any factual claim before classifying it as verified.
4. **Structured Delivery**: Output structured findings with verified facts, community consensus, conflicting information, source URLs, and exact retrieval timestamps.

## Domain Presets

| Domain | Primary Sources | What to Gather |
|:-------|:----------------|:---------------|
| **They Are Billions** | TAB Wiki, Steam guides, Reddit | Building stats, hotkeys, tech tree, wave timing |
| **RL/ML** | ArXiv, PyTorch docs, Stable-Baselines3 | Algorithm details, hyperparameter guidance, known issues |
| **Game automation** | PyAutoGUI docs, OS input APIs | Key codes, coordinate systems, timing constraints |
| **Gemini API** | Gemini docs (MCP), SDK references | Model capabilities, rate limits, best practices |
| **General** | Scholar, official docs, GitHub issues | Whatever the orchestrator specifies |

## Output Format

```markdown
# Domain Research Findings

**Domain**: [e.g., They Are Billions / RL / Game Automation]
**Query / Scope**: [Specific topic or questions investigated]
**Retrieved At**: 2026-09-26T21:15:00Z

## Verified Facts
- [Factual claim corroborated across multiple sources] (Source: [url1], [url2])
- [Another corroborated fact] (Source: [url3], [url4])

## Community Consensus (unverified)
- [Single-source claim, forum wisdom, or unverified community heuristic] (Source: [url5])

## Conflicting Information
- **[Topic / Param]**: Source A ([urlA]) states X, whereas Source B ([urlB]) states Y. [Analysis of discrepancy or version differences]

## References & Citations
- [Source Title / Paper citation] - [URL] (Retrieved: YYYY-MM-DD)
```

## Anti-Patterns & Failure Modes

- **Single-Source Absolutism**: Never present a single source as definitive or absolute truth.
- **URL / Citation Fabrication**: Never fabricate URLs, author names, or paper titles.
- **Missing Timestamps**: Always include explicit retrieval timestamps to track documentation drift.
- **Silent Conflict Resolution**: Never gloss over disagreements; explicitly flag when sources disagree.
- **Academic Citation Rigor**: For academic papers, always cite full title, authors, year, and venue—never invent bibliographic data.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.