skills/bennoloeffler/claude-code-personal-agent-plugins-skills/ai-create-diagram/Gen Agent Trust Hub
ai-create-diagram
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references an installation script at
scripts/create_viz_install.shto set up the D2 environment. While the source D2 (d2lang.com) is a well-known layout tool, the script itself is not provided for review. - [COMMAND_EXECUTION]: The skill executes shell commands, including
mkdirto create output directories andd2to compile diagram files into SVG and PNG formats. - [DYNAMIC_EXECUTION]: The agent dynamically generates
.d2source code at runtime based on user instructions, which is then passed to thed2binary for execution/rendering. - [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user descriptions to generate diagram source code, representing a potential injection surface.
- Ingestion points: User-supplied diagram descriptions via the
<beschreibung>argument. - Boundary markers: None explicitly defined in the instructions to prevent the agent from obeying instructions embedded within the diagram description.
- Capability inventory: File writing, directory creation, and execution of the
d2binary via shell. - Sanitization: No sanitization or validation of the user input is mentioned before it is interpolated into the D2 source code.
Audit Metadata