cloud-s3-exposure

Pass

Audited by Gen Agent Trust Hub on Sep 20, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill identifies and processes data from untrusted external sources which is then used to construct shell commands.
  • Ingestion points: As described in the Method section of SKILL.md, the agent is instructed to extract bucket names and identifiers from asset URLs, JavaScript source code, and DNS CNAME records.
  • Boundary markers: Absent. The instructions provide no delimiters or specific warnings to the agent regarding the handling of these extracted strings.
  • Capability inventory: The skill utilizes powerful CLI tools including awscli and gsutil for listing, reading, and writing files (aws s3 ls, aws s3 cp, gsutil ls).
  • Sanitization: Absent. The skill does not provide instructions for validating or escaping the strings extracted from web assets before they are used as arguments in shell commands, creating a surface for command injection.
  • [COMMAND_EXECUTION]: The skill relies on the execution of multiple system-level CLI tools (aws, gsutil, s3scanner, gcpbucketbrute) to perform its primary function. While consistent with the stated bug bounty use case, the direct use of external data in these commands presents an inherent risk.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 20, 2026, 11:01 PM
Security Audit — agent-trust-hub — cloud-s3-exposure