prowler-attack-paths-query

Fail

Audited by Gen Agent Trust Hub on Mar 28, 2026

Risk Level: HIGHEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill fetches attack path definitions from DataDog's pathfinding.cloud and schema documentation from the Cartography project (via CNCF or prowler-cloud repositories). These are well-known and trusted sources for security research data.
  • [COMMAND_EXECUTION]: Employs Bash commands (curl, jq, python3) to download and process JSON data. While piping to an interpreter is a sensitive pattern, here it is used with a static local command (python3 -c) to parse data from trusted URLs rather than executing remote scripts.
  • [PROMPT_INJECTION]: Ingests external JSON data which is then used to generate code and documentation, creating a surface for indirect prompt injection.
  • Ingestion points: Remote JSON file from https://raw.githubusercontent.com/DataDog/pathfinding.cloud/main/docs/paths.json.
  • Boundary markers: Absent; external data is interpolated into query definitions without isolation markers.
  • Capability inventory: The skill utilizes Edit and Write tools to modify project files in the api/attack_paths/queries/ directory.
  • Sanitization: No validation or sanitization of the external JSON content is described before it is used to generate Python code.
Recommendations
  • HIGH: Downloads and executes remote code from: https://raw.githubusercontent.com/DataDog/pathfinding.cloud/main/docs/paths.json - DO NOT USE without thorough review
Audit Metadata
Risk Level
HIGH
Analyzed
Mar 28, 2026, 01:21 AM