Simple Claude Deep Research Agent
PendingAudited by VirusTotal on May 11, 2026.
Overview
No VirusTotal analysis has been recorded yet. File reputation checks will appear here once the artifact hash has been scanned.
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.
The agent may try to access or render content that a website intended to restrict, potentially violating access rules or user expectations.
The subagent is told to automatically use browser automation when login walls or paywalls might be bypassed, which is not clearly authorized by the research purpose.
"Login walls or paywalls that might be bypassed by rendering" ... "If any of these occur, IMMEDIATELY use Playwright MCP"
Remove the paywall/login-bypass language and require explicit user authorization; instruct the agent to respect paywalls, login requirements, robots policies, and site terms.
If these settings are honored, the agent could modify local project files while performing research, not just return a report in chat.
The local Claude settings grant Edit permission without a visible path or file-type restriction, even though the research workflow mainly needs web access and report generation.
"allow": ["Bash(dir /a C:UsersGZDesktopsimple_deep_research)", "Skill(deep-research)", "WebSearch", "Edit"]
Remove broad Edit permission or restrict it to a dedicated output directory such as research_outputs/, with confirmation before overwriting files.
Sensitive details included in a research request could be shared across multiple agent contexts during the task.
The skill intentionally delegates research prompts and context to multiple general-purpose subagents, which is purpose-aligned but means user-provided details may be copied into subagent prompts.
Use the `Task` tool to launch research subagents with the `general-purpose` subagent_type ... Deploy multiple subagents SIMULTANEOUSLY
Avoid putting private credentials, confidential documents, or sensitive personal data into research prompts; keep subagent prompts narrowly scoped.
It is harder to independently verify the skill’s origin, maintenance status, or intended distribution channel.
The package provenance is limited, so users have less context for who maintains it or where to verify updates.
Source: unknown; Homepage: none
Install only if you trust the publisher; prefer a repository or homepage with review history and pin any external MCP dependencies you choose to add.