Security audit

Pitch

Security checks across malware telemetry and agentic risk

Overview

This pitch-coaching skill is mostly local and disclosed, but it can generate investor-facing pitches with fabricated business facts and silently keeps meeting details on disk.

Review generated pitches carefully and replace any traction, market, team, or fundraising claims with verified facts before using them. Install only if you are comfortable storing pitch foundations, objections, and meeting follow-up metadata locally under ~/.openclaw/workspace/memory/pitch, and periodically review or delete those files.

SkillSpector

By NVIDIA

Vulnerability Patterns

Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
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 (5)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 95% confidence
Finding: The skill directs the agent to use local scripts and persist user business materials under `memory/pitch/`, which implies file read/write capability, but no permissions are explicitly declared. This creates a mismatch between the skill's documented behavior and its declared security model, increasing the risk of unauthorized file access, unexpected data retention, or abuse if the agent/runtime does not enforce least privilege.

Description-Behavior Mismatch

Medium

Confidence: 96% confidence
Finding: The script presents fabricated traction, market size, team pedigree, and fundraising details as if they were generated from the user's stored foundation data. In a pitch-coaching skill, this is dangerous because users may unknowingly repeat false business claims to investors, customers, or partners, creating fraud, reputational, legal, and trust risks.

Description-Behavior Mismatch

Medium

Confidence: 94% confidence
Finding: The script stores company/contact meeting metadata to a persistent file under the user's home directory even though its stated purpose is drafting follow-up emails, not maintaining a CRM-like history. This creates undisclosed retention of potentially sensitive business relationship data, which can expose meeting history, next steps, and discussion topics to other local processes, users, or future components that read this memory.

Context-Inappropriate Capability

Medium

Confidence: 89% confidence
Finding: The code establishes a local meeting-history storage mechanism in ~/.openclaw/workspace/memory/pitch, giving the skill a durable memory capability beyond one-off advisory output. In the context of a pitch-coaching skill, this is more dangerous because users may reasonably expect ephemeral drafting help, not background accumulation of investor, customer, or partner interaction records.

Missing User Warnings

Medium

Confidence: 96% confidence
Finding: Meeting details are written to disk without any user-facing disclosure, confirmation, or ability to decline persistence. Because the saved fields include company name, meeting type, key discussion point, and next step, the script silently captures commercially sensitive context that could be accessed later without the user's awareness.

VirusTotal

56/56 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.