atlas-cloud

Pass

Audited by Gen Agent Trust Hub on Nov 14, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill facilitates the processing of untrusted data, specifically text prompts and media attachments, which are sent to external Large Language Models (LLMs) and generative media models. This represents an inherent attack surface for indirect prompt injection.
  • Ingestion points: Untrusted data enters the agent's context through the prompt, reference_images, reference_audios, image_paths, and image_urls parameters defined in SKILL.md.
  • Boundary markers: The instructions do not specify boundary markers or explicit delimiters to prevent the agent from obeying instructions embedded in the input data.
  • Capability inventory: The skill uses tool.execute to perform network operations to the Atlas Cloud API and manages file read/write operations within the projects/ directory as described in SKILL.md.
  • Sanitization: There is no mention of sanitization or validation logic for the content of prompts or media files before they are processed by the tool.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 14, 2026, 08:42 AM