Research Assistant Deep
PassAudited by ClawScan on May 4, 2026.
Overview
This is a coherent web-research skill that uses disclosed search/fetch tools and a simple local URL scoring helper, with no evidence of hidden, destructive, credential, or persistence behavior.
This skill appears safe for normal research use. Before installing, be aware that research queries and URLs may be sent through web search/fetch services, and verify important conclusions from the cited sources.
Findings (2)
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.
If a user researches private or sensitive topics, those query terms or URLs may be sent to the search/fetch provider.
The skill depends on external web search and fetch tooling. This is disclosed and central to the skill, but it means user research terms and requested URLs may leave the local agent context.
requires:\n tools: ['web_search', 'web_fetch']
Avoid including secrets or unnecessary personal data in research queries, and understand that web search/fetch tools may involve external services.
Misleading or manipulative web pages could influence the generated report if not checked against other sources.
The skill brings retrieved web page content into the agent's working context. That is expected for research, and the workflow mitigates it by requiring cross-referencing, contradictions, confidence levels, and citations.
For each promising result, use `web_fetch` with `extractMode: "markdown"` and `maxChars: 8000`.
Treat fetched pages as evidence, not as instructions; verify important claims against cited sources before relying on the report.