canvas-layout

Warn

Audited by Gen Agent Trust Hub on Apr 10, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill constructs shell commands by interpolating variables like [canvas_path] and [options] directly into the command string (python3 scripts/canvas_layout.py [canvas_path] [algorithm] [options]). This presents a command injection risk if these variables contain shell metacharacters such as semicolons or pipes. Evidence: Workflow steps 4 and 6.
  • [PROMPT_INJECTION]: The skill has a surface for indirect prompt injection because the agent is directed to analyze untrusted canvas content to determine layout strategies for complex files. Evidence: 'For Complex Layouts (30+ Nodes)' section.
  • Ingestion points: Target Obsidian canvas files and local reference markdown files (../canvas/references/).
  • Boundary markers: Absent; the agent reads content without delimiters to separate data from instructions.
  • Capability inventory: Execution of shell commands via python3 for file manipulation.
  • Sanitization: No input validation or sanitization is mentioned for the canvas data or command arguments.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Apr 10, 2026, 05:07 PM