hardcoded-credential-hunt

Fail

Audited by Gen Agent Trust Hub on Jul 31, 2026

Risk Level: HIGHCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes shell commands and inline Python scripts to automate security reconnaissance. It uses curl to retrieve web content and python3 -c to execute local logic that parses the response for secrets.
  • [EXTERNAL_DOWNLOADS]: The skill retrieves data from external, user-defined target URLs. This is the primary function of the tool but constitutes a point of entry for untrusted data.
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it processes untrusted data from external websites.
  • Ingestion points: Content fetched via curl from https://target.com (SKILL.md).
  • Boundary markers: None identified in the skill instructions.
  • Capability inventory: The skill uses terminal to run curl, grep, bash, and python3 (SKILL.md).
  • Sanitization: Basic JSON parsing and regex matching are used for extraction, but these do not prevent the agent from being influenced by instructions embedded within the analyzed data.
  • [SAFE]: The automated scan findings for remote code execution are false positives. The reported patterns involve piping web content into a local Python process with a hardcoded script string (python3 -c "...") for data analysis, which does not involve executing the remote content itself.
Recommendations
  • HIGH: Downloads and executes remote code from: https://target.com/PATH, https://target.com:PORT/ERROR_TRIGGER_PATH, https://target.com/api/config - DO NOT USE without thorough review
Audit Metadata
Risk Level
HIGH
Analyzed
Jul 31, 2026, 04:47 PM
Security Audit — agent-trust-hub — hardcoded-credential-hunt