Security audit

research-gif-enricher

Security checks across malware telemetry and agentic risk

Overview

This skill matches its stated Bear-note GIF-enrichment purpose, but it can batch rewrite private notes without preview or rollback and may send note-derived search terms to an external GIF service.

Install only if you are comfortable granting read/write access to Bear notes tagged 「待整理」 and sharing note-derived GIF search terms externally. Back up or export the affected Bear notes first, review the tagged notes, and prefer adding a dry-run or confirmation step before running the batch script.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (5)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 94% confidence
Finding: The skill clearly invokes shell commands and external/network-capable tooling (`grizzly`, `gifgrep`, and fallback `web_search`/`web_fetch`) while declaring no explicit permissions or warning boundaries. That creates a transparency and policy-enforcement gap: users and the platform cannot accurately assess or constrain that the skill reads local notes, modifies them, and may send note-derived queries off-device.

Missing User Warnings

Medium

Confidence: 91% confidence
Finding: This skill performs bulk, destructive content changes by appending media and removing a workflow tag from Bear notes, yet it does not prominently warn the user about those modifications. In a note-management context, silent bulk edits can cause data integrity issues, workflow disruption, and accidental loss of organizational metadata across many notes.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The fallback to `web_search`/`web_fetch` can transmit keywords derived from private note titles or contents to external services, but the skill gives no privacy warning or minimization guidance. In a research-notes context, even short derived queries may expose confidential topics, projects, names, or sensitive themes to third parties.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The script automatically rewrites note contents to remove the tag and replaces the full body without any confirmation, preview, or rollback. In a note-taking skill, silent destructive modification is risky because it can remove user metadata or formatting unexpectedly and overwrite concurrent edits.

Missing User Warnings

Medium

Confidence: 97% confidence
Finding: The script derives a search query from note titles and sends it to Tenor's external API, which leaks note-derived metadata to a third party. Even though only the title-derived query is sent rather than full note text, research note titles can contain sensitive topics, project names, or personal information.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.