ralphthon-seoul-mid-presentation-ko

Warn

Audited by Gen Agent Trust Hub on Mar 29, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs the user or agent to install the slides-grab package from the npm registry and use playwright to download and install the Chromium browser engine.
  • [COMMAND_EXECUTION]: The workflow involves several shell commands including npm install, npx playwright install, and various slides-grab CLI operations such as validate, build-viewer, and edit. These commands are executed based on the logic defined in SKILL.md.
  • [REMOTE_CODE_EXECUTION]: The skill contains instructions to register the downloaded slides-grab package as a persistent agent capability using npx skills add ./node_modules/slides-grab -g -a codex --yes --copy. This pattern effectively installs and authorizes third-party code to run with the agent's permissions.
  • [PROMPT_INJECTION]: The workflow processes untrusted user input from interview responses to generate slide content and file structures. Because the skill lacks explicit boundary markers or sanitization for this input, it presents an indirect prompt injection surface where malicious user data could attempt to influence the slide generation or validation tools.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 29, 2026, 02:53 AM