slides-generator

Fail

Audited by Gen Agent Trust Hub on Feb 21, 2026

Risk Level: CRITICALREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • REMOTE_CODE_EXECUTION (CRITICAL): The skill uses the bash <(curl ...) pattern to download and execute a script from https://raw.githubusercontent.com/hummingbot/.... As 'hummingbot' is not a trusted organization, this allows for arbitrary code execution by the script maintainer.
  • COMMAND_EXECUTION (HIGH): The skill requires the agent to execute multiple shell commands, including global package installations and writing user-controlled content to files in /tmp.
  • EXTERNAL_DOWNLOADS (MEDIUM): The skill initiates downloads and installations for the Python package fpdf2 and the Node.js package @mermaid-js/mermaid-cli without version pinning or integrity validation.
  • PROMPT_INJECTION (LOW): The skill presents an indirect prompt injection surface. Evidence: 1. Ingestion points: Markdown content provided by users in Step 3. 2. Boundary markers: Uses 'SLIDES_EOF' heredoc, but lacks explicit 'ignore embedded instructions' warnings for the processing agent. 3. Capability inventory: Execution of bash commands and network operations via curl. 4. Sanitization: No sanitization of user-provided markdown content is performed before processing.
Recommendations
  • HIGH: Downloads and executes remote code from: https://raw.githubusercontent.com/hummingbot/skills/main/skills/slides-generator/scripts/generate_slides.sh - DO NOT USE without thorough review
  • AI detected serious security threats
Audit Metadata
Risk Level
CRITICAL
Analyzed
Feb 21, 2026, 09:40 AM