componentize-design

Pass

Audited by Gen Agent Trust Hub on Jul 22, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes design studies (HTML files) from the 'docs/designs/' directory, which creates a surface for indirect prompt injection where malicious instructions embedded in design files could influence the agent's behavior during code generation. 1. Ingestion points: Design studies located in the 'docs/designs/' directory. 2. Boundary markers: Absent for the processed design content. 3. Capability inventory: The agent performs file writes to create token contracts and component files, and executes repository linting and testing commands. 4. Sanitization: No specific sanitization or filtering of the HTML content is described beyond targeted value extraction.
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands to extract data from design files and verify the generated output. Evidence: Use of 'grep' to extract property values from design studies in Step 1, and execution of the repository's lint and test commands in Step 5.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 22, 2026, 01:51 AM
Security Audit — agent-trust-hub — componentize-design