stelow-product-execution-critique
Pass
Audited by Gen Agent Trust Hub on Aug 14, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill audits external codebases and URLs, processing potentially untrusted content through the LLM. Ingestion points include spec-tech files and URL snapshots via agent_browser. There are no explicit boundary markers to isolate untrusted data, and no sanitization is performed.
- [REMOTE_CODE_EXECUTION]: The skill executes embedded Node.js and Python code snippets using shell flags. Node.js is used to update the stelow.json tracking file, and Python is used to process entity graph data from the sem tool.
- [COMMAND_EXECUTION]: The skill relies on various system tools like git, sem, and cymbal for change detection and auditing tasks.
- [EXTERNAL_DOWNLOADS]: The skill utilizes a browser tool to visit remote websites during the audit process.
Audit Metadata