ClawHub
Back to skill

Security audit

Academic Deep Research Pro

Security checks for vulnerabilities and agentic risk

Overview

This is a disclosed research-workflow skill that uses web search, web fetch, and optional parallel sub-agents after user planning approval, with no executable code or persistence found.

Install this if you want a rigorous, multi-step research workflow and are comfortable with it performing web searches, fetching pages, and optionally spawning parallel research agents after you approve a plan. Avoid using it for sensitive private topics unless you are comfortable with the research topic and search terms being sent to external web tools.

A.I.G

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (5)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The README advertises broad activation phrases such as asking for "deep research on..." or "exhaustive analysis of...", which are generic requests that may unintentionally trigger this skill when a user did not explicitly intend to invoke it. In an agent ecosystem, overly broad natural-language triggers can cause misrouting, unexpected tool use, and execution of a heavyweight research workflow on sensitive or unrelated prompts.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The trigger guidance is very broad and can cause this skill to activate for a wide range of ordinary research-related requests, increasing the chance that it overrides more specific or safer skills. Because the skill mandates extensive web research, multi-step workflows, and session spawning, over-triggering can unnecessarily expand tool use, data exposure, and operational complexity.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
93% confidence
Finding

The workflow explicitly labels Phase 3 as 'Auto-Execute' and proceeds to run research cycles using external tools without an explicit per-action approval boundary. In an agent setting, this creates autonomous tool use based on interpreted user goals, which can lead to unintended external queries, excessive data retrieval, or processing of sensitive topics without a fresh confirmation step.

Content

Scanner excerpt · example.md (reported line 76)May include surrounding context.

md
---

## Phase 3: Research Cycles (Auto-Execute)

### Theme 1: Market Landscape — Cycle 1

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The invocation guidance allows activation by a very broad natural-language phrase ("deep research" / "exhaustive analysis") without requiring explicit confirmation that the user wants the high-autonomy research workflow. In practice, this can cause accidental triggering of a tool-using skill in ordinary conversation, increasing the chance of unintended web access, scope escalation, or expensive automated behavior.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The document explicitly describes Phase 3 as "fully automated" and directs a repeated sequence of web_search and web_fetch, but it does not warn the user that external network requests will be made on their behalf. This reduces informed consent and can lead to unintended disclosure of user-provided topics, autonomous browsing of untrusted content, and unexpected resource consumption.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.