Academic Deep Research Pro
Security checks across malware telemetry and agentic risk
Overview
This instruction-only research skill is mostly coherent and low-risk, but users should note that it performs external web research and may spawn sub-agents despite an offline/no-cloud claim.
This skill appears safe to install as an instruction-only research workflow. Before using it, remember that it is not truly offline: it relies on web_search and web_fetch, and it may use parallel sub-agents. Review and narrow the research plan before approving automated execution, especially if your topic includes confidential, regulated, or sensitive information.
VirusTotal
VirusTotal findings are pending for this skill version.
Risk analysis
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.
Approving the plan may trigger many web research actions using the approved topic and scope.
After the user approves the plan, the skill directs the agent to run multiple web searches and fetches without intermediate checkpoints. This is expected for exhaustive research, but it is still automated external tool use.
Wait for explicit approval before Phase 3. ... Phase 3: Mandated Research Cycles (NO STOPS) ... `web_search` broad query (count=20) ... `web_fetch` primary sources.
Before approving Phase 3, confirm the research scope, number of themes, source limits, and whether the topic contains sensitive information.
Research details may be processed by additional agent sessions during parallel work.
The skill may delegate parts of the research request to sub-agents. This is disclosed and purpose-aligned, but it means the research prompt and topic can be shared across spawned agent sessions.
Use parallel sub-agents when themes are independent. ... `sessions_spawn(task="Research market landscape with 2 cycles... return findings, gaps, sources")`
Avoid including private or sensitive details in research prompts unless you are comfortable with them being included in delegated sub-agent tasks.
A user might incorrectly believe the skill does not use network-based research tools.
The README's offline/no-cloud claim is inconsistent with the documented reliance on web_search and web_fetch. The web-tool use is disclosed elsewhere, so this appears to be a documentation inconsistency rather than hidden behavior.
Works offline — ✅ Yes ... No API keys, no cloud services ... Phase 3: Execution (Auto) ... `web_search` ... `web_fetch`
Treat the skill as web-connected when using it, and avoid submitting confidential research topics if external web-tool processing is not acceptable.