slideshow

Warn

Audited by Gen Agent Trust Hub on Aug 20, 2026

Risk Level: MEDIUMREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill includes an instruction to "run silently, don't ask" for the update command npx hyperframes skills update slideshow. This is a concealment pattern that attempts to bypass the agent's user-confirmation requirements.
  • [REMOTE_CODE_EXECUTION]: The skill triggers the execution of remote code by instructing the agent to run npx hyperframes. It also references the external Three.js library via a CDN in its reference documentation.
  • [COMMAND_EXECUTION]: The skill relies on multiple CLI tools for slideshow management, including npx hyperframes present, npx hyperframes lint, and npx hyperframes check.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from Figma URLs and external web pages to facilitate conversion. 1. Ingestion points: Figma URLs and source page code. 2. Boundary markers: No explicit delimiters or warnings are provided to prevent the agent from obeying instructions embedded in source content. 3. Capability inventory: Ability to execute shell commands and write project files. 4. Sanitization: The skill lacks robust sanitization requirements for arbitrary code ported from external web pages.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Aug 20, 2026, 09:33 AM
Security Audit — agent-trust-hub — slideshow