Back to skill

Security audit

Cross-Channel Notify

Security checks for vulnerabilities and agentic risk

Overview

The skill has a clear notification purpose, but its documented send workflow uses unsafe shell message construction that could turn crafted message content into local command execution or unintended email headers.

Review before installing or using this skill with untrusted, copied, or third-party message text. It sends content to configured email and iMessage services, and the documented email command pattern should be fixed to avoid shell heredoc injection and email header injection before routine use.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (1)

T09 · Insecure Skill Coding Practices

Error
Location
SKILL.md:18
Finding

User-Controlled Message Content Embedded in a Shell Heredoc

Content
View full analysis
To: Subject: EOF ``` ``` ### Technical Analysis The workflow collects recipient, subject, and body values from user-controlled input and directs the Agent to interpolate them into a shell heredoc. The heredoc uses the fixed delimiter `EOF`. Quoting the delimiter as `'EOF'` disables shell expansion within the heredoc content, but it does not prevent a line containing exactly `EOF` from terminating the heredoc. If an attacker includes such a line in the message body, the resulting generated shell script can close the heredoc prematurely. Text following the injected delimiter is then parsed as shell syntax rather than email content. For example, a malicious body could conceptually contain: ```text Routine notification EOF id > /tmp/heredoc-command-executed # ``` When this content is substituted for ``, the first attacker-controlled `EOF` ends the heredoc. The following command is consequently processed by the shell. The fixed closing delimiter later in the generated command may produce an error, but only after the injected command has had an opportunity to execute. The `To` and `Subject` fields are also inserted into an RFC 5322 message without validation. If either value contains carriage-return o ...[truncated 2016 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (1)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

This skill transmits user-provided message content to two external services, Himalaya for email and BlueBubbles for iMessage, but does not explicitly warn the user that their content and recipient data will leave the local agent environment. That creates a meaningful privacy and consent risk, especially because the skill is designed for multi-channel broadcast, which increases the chance of unintentionally sharing sensitive information more broadly than the user expects.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.