search-harvester

• Outcome: Review
• Latest audit: Aug 7, 2026 · 10:40 AM UTC
• Version: 1.0.0

Overview

This skill openly routes search scraping through Tor and optional public proxies to bypass search-engine blocking, so users should review it carefully before installing.

Install only if you intentionally need this Tor-based scraping workflow, have authorization for the search and outreach activity, and accept the privacy and ToS risks. Avoid sensitive or regulated queries, prefer official APIs or approved data sources where available, and run Tor as a constrained unprivileged service rather than copying the sudo setup directly.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands

Findings (13)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 95% confidence
• Finding: The skill performs network access, shell execution, and file writes but does not declare permissions or present clear guardrails. This creates a transparency and control failure: a caller or hosting platform may invoke a capability-rich skill without understanding that it launches Tor, writes temporary files, and reaches external services.

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: The README explicitly promotes routing search queries through rotating Tor exit nodes to avoid search-engine blocking and hide the server IP, but it does not warn users about privacy exposure to exit nodes, potential ToS violations, or legal/operational risks from automated scraping through anonymity infrastructure. In this context, the omission is security-relevant because the skill is designed to evade provider anti-bot controls, which increases the chance of misuse, account/network abuse complaints, and unsafe handling of sensitive queries.

Vague Triggers

Medium

• Confidence: 82% confidence
• Finding: The invocation text is broad enough to trigger on generic research or search-related requests, which increases the chance that the agent routes ordinary searches through Tor/proxies and executes scraping behavior without specific user intent. In this skill, that matters because invocation implies anti-bot evasion, network pivoting, and potentially policy-sensitive automation.

Missing User Warnings

High

• Confidence: 97% confidence
• Finding: The skill routes traffic through Tor exit nodes and optionally public HTTP proxies but does not present a prominent user warning about privacy, trust, legal, and abuse implications. Users may unknowingly send queries and harvested targets through third-party infrastructure, exposing search terms and creating risk of interception, abuse complaints, or policy violations.

Vague Triggers

Medium

• Confidence: 91% confidence
• Finding: The skill description explicitly frames the tool as a fallback whenever search engines block the server IP, which broadens activation from a narrow SEO workflow into any blocked search task. In context, that means the skill is designed to evade search-engine anti-abuse controls by routing requests through Tor and rotating exits, increasing the likelihood of misuse and policy-violating automation.

Vague Triggers

High

• Confidence: 97% confidence
• Finding: The trigger condition 'Any search-engine query that gets blocked from the server IP' is an explicit activation rule for circumvention. This makes the skill more dangerous because it instructs the agent to switch to anonymity infrastructure whenever a provider's abuse defenses engage, effectively operationalizing evasion rather than legitimate search.

Missing User Warnings

Medium

• Confidence: 93% confidence
• Finding: The skill routes traffic through Tor and optionally public proxies while harvesting queries and URLs, but it does not clearly warn that queries, targets, and metadata may be exposed to third-party exit nodes or proxy operators. In a security context, this creates privacy, compliance, and data-handling risk, especially if user queries or downstream targets are sensitive.

Missing User Warnings

Medium

• Confidence: 91% confidence
• Finding: The script sends user-supplied search queries and harvested target URLs to third-party services with no explicit consent or warning, which can leak sensitive research topics, client names, or campaign intent. In this skill's context, the disclosure is more concerning because it is designed to evade search-engine blocking via Tor rotation, increasing the likelihood of high-volume external exposure and policy-violating activity.

Missing User Warnings

Low

• Confidence: 75% confidence
• Finding: The script writes harvested URLs and associated query metadata to disk without warning the user that this may persist potentially sensitive prospecting data or overwrite an existing file. While lower severity than the network disclosure issue, it can still expose client research, outreach targets, or search strategy to other local users or later processes.

Ssd 4

Medium

• Confidence: 98% confidence
• Finding: The skill explicitly frames Tor rotation as a way to ensure the server IP never touches the search engine and to avoid bans, captchas, and bot-blocks. That is not incidental networking; it is operational guidance for bypassing anti-abuse controls and repeatedly obtaining access after blocks.

Ssd 4

Medium

• Confidence: 99% confidence
• Finding: The rotate-and-retry loop instructs the operator to change Tor identity after failures and keep retrying until a clean exit node is found, including guidance to rotate even between successful queries. This is a textbook gradual evasion pattern that increases resilience against anti-bot defenses and materially raises abuse potential.

Sudo/Root Execution

Medium

• Category: Privilege Escalation

Content

#   SOCKSPort 127.0.0.1:19050
#   ControlPort 127.0.0.1:19051
#   DataDirectory /tmp/tor-harvest-data
sudo tor -f /tmp/tor-harvestrc        # run as background process
# Verify listening:
sudo ss -tlnp | grep -E "19050|19051"
# Verify exit IP:

• Confidence: 88% confidence
• Finding: sudo

Sudo/Root Execution

Medium

• Category: Privilege Escalation

Content

#   SOCKSPort 127.0.0.1:19050
#   ControlPort 127.0.0.1:19051
#   DataDirectory /tmp/tor-harvest-data
sudo tor -f /tmp/tor-harvestrc        # run as background process
# Verify listening:
sudo ss -tlnp | grep -E "19050|19051"
# Verify exit IP:

• Confidence: 84% confidence
• Finding: sudo

VirusTotal

VirusTotal findings are pending for this skill version.

Static analysis

No suspicious patterns detected.