ckm-banner-design

Pass

Audited by Gen Agent Trust Hub on Jun 29, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill documents the use of local scripts located in the .claude/skills/ directory, specifically gemini_batch_process.py for AI image generation and screenshot.js for exporting HTML banners to PNG. These commands are integral to the design and export workflow.
  • [EXTERNAL_DOWNLOADS]: The agent is instructed to use a browser to research design inspiration on Pinterest, which involves fetching external content. This is consistent with the skill's purpose for creative research.
  • [DATA_EXPOSURE]: The skill accesses local project files such as docs/brand-guidelines.md to maintain brand consistency and writes generated output to the assets/banners/ directory.
  • [INDIRECT_PROMPT_INJECTION]:
  • Ingestion points: Ingests external design references from Pinterest and brand guidelines from local project files.
  • Boundary markers: No explicit delimiters are used when processing external inspiration or brand data.
  • Capability inventory: Includes execution of local Python/Node.js scripts and file system writes for asset generation.
  • Sanitization: No specific sanitization logic is described for the data retrieved during the research step.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 29, 2026, 08:01 AM
Security Audit — agent-trust-hub — ckm-banner-design