ClawHub
Back to skill

Security audit

Graincrawl

Security checks across malware telemetry and agentic risk

Overview

This skill is a local Granola archive search helper, with some metadata quality problems but no evidence of hidden, destructive, or unrelated behavior.

Before installing, understand that this skill can guide an agent to read and query private Granola notes, transcripts, and panels, and to sync from private API or desktop cache sources when stale or requested. The publisher should fix the frontmatter description and trigger text so users see the real scope at install time.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • 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
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Description-Behavior Mismatch

Medium
Confidence
98% confidence
Finding
The manifest description declares no meaningful user-facing purpose beyond the word "metadata:", while the body documents concrete operational behavior: checking archive freshness, syncing from sources, and querying notes/transcripts/panels. This is a semantic mismatch because the declared description does not reflect the actual skill behavior or scope.

Intent-Code Divergence

Medium
Confidence
97% confidence
Finding
The changelog states that the frontmatter description was corrected into a real trigger, yet the actual frontmatter still shows `description: "metadata:"`. This is an active contradiction between documentation about the file's intent/state and what the file currently contains.

Vague Triggers

Medium
Confidence
96% confidence
Finding
The trigger definition says the skill activates whenever the user requests something that matches the description above, but the description above is effectively just 'metadata:' and does not define a concrete scope. This makes activation boundaries unclear and risks unintended invocation because there is no specific trigger phrase list or meaningful exclusion criteria.

Natural-Language Policy Violations

Medium
Confidence
94% confidence
Finding
The skill instructions are written in Indonesian and present the policy layer as mandatory ('wajib') without offering the user a language or locale choice. This can violate language/locale policy expectations when no user opt-in or justification for the language constraint is provided.

VirusTotal

61/61 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.