diagram-generator
Pass
Audited by Gen Agent Trust Hub on Nov 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted architectural briefs and system descriptions to generate technical diagrams. This constitutes an indirect prompt injection surface. However, the skill incorporates a security-conscious design by requiring a validation pass through a bundled checker script (
scripts/check-diagram.mjs) that explicitly verifies the artifact is self-contained and free of external asset references (E012), which effectively mitigates data exfiltration or remote script loading risks in the generated output. - Ingestion points: User-provided system descriptions are ingested via files like
SKILL.md,AGENTS.md, and various example templates. - Boundary markers: While the prompt logic does not explicitly use strong delimiters, the workflow requires a planning phase (
diagram-ir) that structures the data before rendering. - Capability inventory: The skill has the capability to write HTML files and execute local scripts for auditing purposes.
- Sanitization: The skill uses a post-generation audit script to detect and fail on template placeholders or external resource references.
- [COMMAND_EXECUTION]: The instructions direct the agent to run a local Node.js script (
node scripts/check-diagram.mjs) to validate the geometry and integrity of the generated diagram. This script uses standard Node.js libraries (fs,path) and contains no network operations or dangerous dynamic execution patterns.
Audit Metadata