uzi-skill-stock-analyzer
Warn
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill requires the user or agent to clone source code from an external repository (github.com/wbh604/UZI-Skill) and install multiple Python dependencies (akshare, yfinance, duckduckgo-search) that are not part of the standard agent environment.
- [REMOTE_CODE_EXECUTION]: Installation instructions for platforms like Gemini CLI and Codex involve installing extensions or following guides directly from remote URLs, which executes external code within the agent's environment.
- [COMMAND_EXECUTION]: The skill utilizes subprocess.run to execute internal Python scripts as part of its multi-stage analysis pipeline.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external financial sources and search results, which are then processed by the agent during the quantitative judgment phase (HARD-GATE), creating a potential surface for indirect injection without specified sanitization.
- [DATA_EXFILTRATION]: The skill features a 'remote mode' that establishes a Cloudflare Tunnel (trycloudflare.com) to provide public access to locally generated reports, representing a potential vector for data exposure.
Audit Metadata