Deep Research Agent
PassAudited by VirusTotal on May 3, 2026.
Overview
Type: OpenClaw Skill Name: eric-deep-research-agent Version: 11.0.0 The skill bundle describes a legitimate deep research agent designed to perform multi-phase information gathering and report synthesis. The instructions in SKILL.md and the templates in research_template.md are well-structured and strictly aligned with the stated purpose of academic and market research. The utility scripts (package_skill.py and quick_validate.py) perform standard packaging and validation tasks without any suspicious network activity, obfuscation, or unauthorized file access.
Findings (0)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
A research request may trigger many searches and webpage/PDF reads, which could take time or consume tool/provider quota.
The skill explicitly directs the agent to use external web research tools and parallel execution. This is expected for a deep-research agent, but it can create substantial network/tool activity.
**Tools Used**: `batch_web_search`, `extract_content_from_websites` ... Run independent searches in parallel (up to 10 concurrent queries)
Use the skill for tasks where broad web research is intended, and set limits if you want fewer sources, less concurrency, or a shorter report.
Misleading or manipulative source content could affect the research synthesis if not treated carefully.
The agent relies on retrieved web and PDF content to build its report. Those sources are untrusted inputs, although the skill also includes cross-verification and contradiction handling.
Access each promising URL ... Extract structured information: facts, statistics, quotes, dates, claims ... Parse PDF documents for detailed data
Keep citations visible, verify important claims against primary sources, and ensure retrieved webpage/PDF text is treated as evidence rather than instructions to the agent.
Users have fewer external signals to judge the origin or maintenance practices of the skill.
The package provenance is limited. This does not indicate malicious behavior, especially because no install step or risky dependencies are shown, but it gives users less external context about the publisher.
Source: unknown; Homepage: none
Review the provided SKILL.md and included scripts before installation, and prefer trusted publishers for sensitive research workflows.