underworld-overseer-save-mapper

Warn

Audited by Gen Agent Trust Hub on Jul 30, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The scripts/setup.sh script performs a git clone from a remote repository (github.com/RobThePCGuy/Underworld-Overseer-Save-Mapper) that is not associated with a verified service or the skill author.
  • [REMOTE_CODE_EXECUTION]: The skill instructions direct the agent to execute the main.py script contained within the third-party repository after the installation process is complete.
  • [COMMAND_EXECUTION]: The skill uses shell commands in scripts/setup.sh to install environment dependencies using pip and manage the repository clone.
  • [PROMPT_INJECTION]: The skill processes untrusted JSON data from game save files, creating a surface for indirect prompt injection.
  • Ingestion points: The scripts/validate_save.py script and the external main.py read user-provided JSON save files from the local filesystem.
  • Boundary markers: The SKILL.md instructions include a mandatory pre-validation step using scripts/validate_save.py to identify structural anomalies before the data is passed to the renderer.
  • Capability inventory: The agent has the capability to execute shell commands (SKILL.md), clone repositories (scripts/setup.sh), and run Python scripts (scripts/validate_save.py).
  • Sanitization: The scripts/validate_save.py script performs type checking, ensuring coordinates are integers and identifying duplicate entries, which helps mitigate basic data malformation attacks.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 30, 2026, 01:50 AM
Security Audit — agent-trust-hub — underworld-overseer-save-mapper