pinyin
Warn
Audited by Gen Agent Trust Hub on Jun 28, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches a compiled binary (
cyeam) from the author's GitHub repository (mnhkahn/cyeam-cli) during the installation phase. - [COMMAND_EXECUTION]: The installation process executes several shell commands, including
curl,tar,chmod, andsudo mv. The skill's primary functionality relies on executing thecyeambinary locally. - [PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection because it processes untrusted user input to generate data that the agent then interprets (JSON) or provides to the user (PDF).
- Ingestion points: Chinese text provided as a CLI argument in
SKILL.md. - Boundary markers: Absent; user text is interpolated directly into command arguments.
- Capability inventory: Execution of local binaries and file-write access for PDF output.
- Sanitization: No evidence of input escaping or validation before processing.
Audit Metadata