case-study

Warn

Audited by Gen Agent Trust Hub on Mar 1, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill dynamically generates a Node.js script (generate-case-study.js) at runtime to build the PPTX file. This script is then executed via the command line using node. This pattern is inherently risky as it executes code constructed from potentially untrusted inputs.
  • [COMMAND_EXECUTION]: The skill executes multiple shell commands to install dependencies (npm install, pip install) and run diagnostic tools (python -m markitdown). While the packages are well-known, the broad capability to run shell commands increases the attack surface.
  • [PROMPT_INJECTION]: The skill is highly vulnerable to indirect prompt injection. It ingests large amounts of untrusted data from user-provided transcripts, notes, and website fetches. This content is used to fill a narrative template and directly influences the generation of the Node.js script in Phase 5, which could lead to code injection if the LLM does not properly sanitize the extracted data.
  • [EXTERNAL_DOWNLOADS]: The skill uses WebFetch or browser automation to retrieve HTML and CSS from user-provided URLs. This functionality could be exploited for Server-Side Request Forgery (SSRF) if an attacker provides a URL targeting sensitive internal network resources.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 1, 2026, 03:28 PM