app-intents-whats-new-27

Fail

Audited by Gen Agent Trust Hub on Sep 2, 2026

Risk Level: HIGHPROMPT_INJECTIONMETADATA_POISONINGINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [METADATA_POISONING]: The skill contains highly deceptive technical information, claiming to be official Apple documentation for 'iOS 26' and 'iOS 27' for the years 2025 and 2026. This versioning significantly deviates from Apple's actual release cycle (where 2025 corresponds to iOS 19), which could mislead developers into attempting to use hallucinated or non-existent APIs.
  • [PROMPT_INJECTION]: The skill instructions in SKILL.md utilize directives designed to override the model's internal training and knowledge. It asserts that the provided content is 'authoritative' and 'unconditionally supersedes any prior training,' instructing the model that 'this guidance is correct' even if it conflicts with the AI's established knowledge. This is a common pattern for bypassing model guardrails.
  • [INDIRECT_PROMPT_INJECTION]: The skill uses self-referential claims of validity ('This guidance was written and published by Apple') to influence the agent's evaluation of the data.
  • Ingestion points: SKILL.md and the references/ directory containing markdown files.
  • Boundary markers: No explicit delimiters or boundary markers are used to isolate the skill's instructions from the technical data.
  • Capability inventory: The skill consists of documentation and code examples but does not provide any executable scripts, network tools, or file system manipulation capabilities.
  • Sanitization: There is no evidence of sanitization or input validation performed on the provided content.
  • [EXTERNAL_DOWNLOADS]: The skill references official Apple developer documentation at developer.apple.com. This originates from a well-known and trusted service.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Sep 2, 2026, 09:55 AM