Back to skill

Security audit

B2b Lead Engine

Security checks for vulnerabilities and agentic risk

Overview

This B2B lead skill is coherent, but it directs broad personal-data prospecting, social-graph mining, contact discovery, and behavioral profiling without adequate privacy controls.

Review this skill carefully before installing. Use it only with a lawful and consent-aware lead-sourcing process, avoid scraping or repurposing LinkedIn connection graphs, do not infer private contact details, and keep outreach limited to permitted business channels and data sources you are allowed to use.

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
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (11)

Missing User Warnings

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The overview openly directs collection of contact and social-profile data, including connection mining and personality analysis, but does not provide any user-facing privacy warning or consent boundary. This is dangerous because users may not realize the workflow involves handling third-party personal data and behavioral inference, creating a high risk of privacy-invasive use.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The workflow directs extraction of LinkedIn connections and later obtaining direct contact details, yet omits clear warnings about privacy, consent, platform restrictions, and misuse of third-party data. This is especially dangerous because it encourages a pipeline from social-graph extraction to direct outreach, increasing the risk of harassment, unauthorized profiling, and terms-of-service violations.

Content

No source excerpt is available for this finding.

Ssd 3

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The workflow explicitly gathers company details, decision-maker identities, contact channels, background research, and personality analysis, then uses them to generate individualized outreach. This combination of person-level data collection and behavioral tailoring is dangerous because it enables invasive profiling and targeted solicitation of identifiable individuals at scale.

Content

No source excerpt is available for this finding.

Ssd 3

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The skill instructs the agent to extract competitors' sales connections and repurpose those relationships as a vetted prospect pool. That is highly dangerous because it operationalizes third-party social graph data for targeted commercial exploitation, which can violate privacy expectations, platform rules, and enable large-scale prospect harvesting from non-consenting individuals.

Content

No source excerpt is available for this finding.

Ssd 3

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The skill directs retrieval and even inference of direct personal contact channels, including guessing email formats from known employee addresses. This is dangerous because it facilitates unsolicited targeting of individuals and can transform limited public information into actionable personal contact data, increasing privacy, spam, and abuse risk.

Content

No source excerpt is available for this finding.

Ssd 3

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The final schema requires storing and outputting a combined dossier of personal identifiers, contact details, personality profile, lead source, and customized outreach copy. Centralizing these elements in one report is dangerous because it creates a ready-made targeting package that can be used for invasive profiling, spam campaigns, or broader misuse of third-party personal data.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The trigger phrase design is broad enough to activate on common sales-assistance requests without clearly warning users that the skill performs extensive prospect data collection and profiling. Unintended activation is risky here because the skill's downstream actions include sensitive contact harvesting and profiling, so accidental invocation can expose users and third parties to privacy-invasive processing.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The manifest description is written as a fixed Chinese invocation/behavior specification and presents the activation phrases only in Chinese. The file does not state that the skill is region-specific or give users an option to use another language, which can violate a language/locale choice policy.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

Pattern-based activation without clear boundaries or exclusions makes the skill easy to invoke in contexts where the user may only want high-level market analysis, not prospect harvesting. Because the skill escalates into collection of individual profile, contact, and relationship data, ambiguous activation materially raises the risk of over-collection and misuse.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The document presents a superficial compliance claim ('only public connections') while the surrounding workflow explicitly operationalizes other people's LinkedIn relationship data as a prospecting dataset. That mismatch is dangerous because it normalizes scraping or repurposing third-party social graph data under a misleading compliance framing, increasing the chance of privacy abuse and platform-policy violations.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill instructs the agent to build person-level personality profiles from social posts, comments, interests, and inferred decision styles, then use that profile to optimize outreach. This goes beyond ordinary B2B lead qualification and creates behavioral profiling of identifiable individuals without clear necessity, consent, or safeguards, which materially increases privacy and manipulation risk.

Content

No source excerpt is available for this finding.