debug
Pass
Audited by Gen Agent Trust Hub on Apr 5, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTIONCREDENTIALS_UNSAFE
Full Analysis
- [COMMAND_EXECUTION]: The workflow requires executing shell commands and scripts to reproduce issues and run experiments (Steps 2 and 6). Although the skill instructs the agent to read user-provided code before execution, this capability represents a security surface.
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes untrusted data from error logs and stack traces (Step 1). Ingestion points include logs and stack traces. Boundary markers are present in the form of instructions to treat logs as data, not instructions. Capabilities include shell execution and file reading. Sanitization involves reading content and obtaining user confirmation before modification.
- [CREDENTIALS_UNSAFE]: The skill instructions permit the agent to read configuration files such as .env (Rules). While it attempts to limit this to keys only, the access to sensitive file paths is a vulnerability.
Audit Metadata