Back to skill

Security audit

表情包 · Emoji Sticker

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marked this skill suspicious because it reports an unpinned installer and external text checking without a clear disclosure; the scan was partial, so its evidence is incomplete.

Consider whether you are comfortable with an installer whose version can change and with text potentially being sent to an external checking service. The scan covered 15 of 16 components and was partial, including malformed manifest data and unresolved references, so the findings do not provide a complete assessment.

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
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • 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 (6)

MCP server referenced without pinned version: 'npx clawhub'.

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The installation command executes clawhub@latest, so the executed package version can change without review. A compromised upstream release could execute code with the installing user's privileges; this is a conditional supply-chain risk, not evidence of a malicious package or an MCP server.

Content

No source excerpt is available for this finding.

The documented trigger expansion includes the overly broad bare keyword “特效” (“special effects”), which can activate the skill outside sticker-generation tasks.

Medium
Category
Trigger Abuse
Confidence
93% confidence
Finding

The changelog explicitly identifies bare words, including “特效,” as added triggers. That term also applies to video editing and other visual-effects tasks, and the entry provides no sticker-specific qualification or exclusion conditions.

Content

No source excerpt is available for this finding.

Mandatory live copy checking lacks a warning that user-provided text is sent to an external API.

Medium
Category
Output Handling
Confidence
96% confidence
Finding

The workflow automatically requires live checking when the companion skill is installed, without explaining the external recipient or requiring authorization to transmit drafts. Although the text is intended for eventual publication, unpublished names and descriptions may still be confidential, and eventual publication does not authorize disclosure to an unspecified checker.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx clawhub'.

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The documented npx command executes the mutable clawhub@latest release rather than a reviewed fixed version. If an upstream release is compromised, following this installation guidance could execute attacker-controlled code with the user's privileges; the source describes a CLI installer, not an MCP server.

Content

No source excerpt is available for this finding.

The skill instructs agents to use an external sensitive-word API without warning that submitted content leaves the local environment.

Medium
Category
Output Handling
Confidence
94% confidence
Finding

The reference directs agents to call a locally installed companion skill that forwards text to a sensitive-word API, without a disclosure or authorization boundary for that transfer. Calling the skill locally does not keep processing local, so unpublished or confidential copy could leave the environment unexpectedly.

Content

No source excerpt is available for this finding.

The documented 8–24-image requirement is not enforced: packs with 2–7 main images are reported as compliant.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The main-image specification sets the count range to 1–24, so the range branch accepts 2–7 images and incorrectly labels that count as 8–24 compliant. Such packs can pass the count check and potentially the overall validation gate, undermining this skill's core compliance guarantee and causing rejected submissions rather than a demonstrated system compromise.

Content

No source excerpt is available for this finding.