skills/varghq/skills/gpt-image-2-6/Gen Agent Trust Hub

gpt-image-2-6

Pass

Audited by Gen Agent Trust Hub on Nov 14, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to process user-supplied text and images, which introduces a surface for indirect prompt injection.
  • Ingestion points: User-provided prompts and local image files are interpolated into shell commands and API payloads, as seen in the recipes in references/recipes.md and the REST call examples in SKILL.md.
  • Boundary markers: The skill does not implement or instruct the agent to use specific delimiters or protective instructions to distinguish untrusted user data from the skill's operational instructions.
  • Capability inventory: The skill utilizes Bash(curl:*) for network operations and Read tools to access local files for processing.
  • Sanitization: There are no explicit instructions for the agent to sanitize, validate, or escape user-supplied strings or file content before including them in the curl commands.
  • [COMMAND_EXECUTION]: The skill provides technical instructions and examples for using shell commands to interact with the varg API.
  • SKILL.md includes curl and jq snippets intended for execution. These commands involve the agent handling local file paths and network communication, which necessitates careful oversight during execution to prevent command injection from malicious user input.
  • [DATA_EXFILTRATION]: The skill contains instructions for uploading local image files to the vendor's API endpoint (api.varg.ai).
  • This data movement is documented as a core feature for image editing and generation tasks. The destination is the official infrastructure of the skill's vendor (vargHQ), representing expected functionality rather than a covert threat.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 14, 2026, 09:49 PM