b2b-sales-prospecting-agent

• Outcome: Review
• Latest audit: May 28, 2026 at 11:18 AM UTC
• Version:
  1.0.0

Overview

This skill appears purpose-built for B2B prospecting, but it deserves review because it handles API credentials, bulk personal contact data, and outbound query metadata with some unclear provenance and consent boundaries.

Install only if you trust the publisher despite the unofficial/author mismatch, have an Explorium API key you are comfortable using here, and have a lawful basis and company policy approval to collect/export prospect emails or phone numbers. Avoid putting confidential GTM strategy or customer names in prompts unless you are comfortable sending that text to the vendor API, and clean up /tmp result files plus exported CSVs when they are no longer needed.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (8)

Context-Inappropriate Capability

Medium

• Confidence: 94% confidence
• Finding: The skill persists an API credential to ~/.agentsource/config.json and also allows reading arbitrary local CSV paths and writing CSV output to arbitrary user-supplied paths. While these features support the tool's workflow, they expand the file-system access surface beyond simple prospect search and can expose local data or leave sensitive credentials/results resident on disk.

Context-Inappropriate Capability

Medium

• Confidence: 97% confidence
• Finding: The optional request_context includes --call-reasoning, which is defined as the user query text and is sent to api.explorium.ai. Free-form prompts often contain sensitive business context, target lists, or internal strategy, so this creates an outbound disclosure channel to a third party.

Vague Triggers

Medium

• Confidence: 90% confidence
• Finding: The trigger phrases are broad and generic for normal business conversations, such as 'build a list of' and 'search for companies,' which can cause the skill to activate in contexts where the user did not intend prospecting automation or external data lookups. Unintended invocation matters here because the skill can perform shell, file, and network operations and may initiate remote searches involving user-provided business criteria.

Missing User Warnings

Medium

• Confidence: 88% confidence
• Finding: The skill promotes retrieval and CSV export of verified emails and phone numbers without a clear user-facing privacy notice, lawful-use warning, or handling guidance for personal contact data. In context, this is more sensitive because the workflow is explicitly designed for bulk lead generation and enrichment, which can implicate privacy, consent, and downstream misuse risks.

Missing User Warnings

Medium

• Confidence: 90% confidence
• Finding: The example explicitly exports a list of named decision-makers with verified email addresses to a local CSV without any warning about privacy, lawful basis, retention, or secure handling. In a prospecting skill, this increases the chance that users will casually collect, store, and redistribute personal contact data in ways that violate internal policy, privacy law, or lead to unintended exposure on shared devices.

Missing User Warnings

Medium

• Confidence: 90% confidence
• Finding: The example explicitly instructs the agent to enrich and export CTO names, emails, and phone numbers without any privacy notice, consent check, lawful-basis guidance, or usage constraints. In a lead-generation skill whose core purpose is finding decision-makers, this omission increases the chance of privacy-law violations, unauthorized processing of personal data, and misuse of direct contact details for unsolicited outreach.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The reference explicitly documents enrichment features that return personal email addresses and direct/mobile phone numbers, but it provides no privacy, consent, lawful-basis, or acceptable-use guidance. In a prospecting skill whose purpose is outbound lead generation, this omission materially increases the risk of misuse, regulatory noncompliance, and unauthorized processing of personal data at scale.

Ssd 3

Medium

• Confidence: 97% confidence
• Finding: The helper builds request metadata that can include natural-language user query text and forwards it to the vendor API. In a sales-prospecting context, those queries may contain confidential go-to-market plans, customer names, or targeting rationale, so transmitting them increases privacy and data-governance risk.

VirusTotal

64/64 vendors flagged this skill as clean.