The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the workspace to generate visual content. This includes brand guidelines, creative briefs, and copy scripts which are interpolated into prompts for AI image generation and into HTML design templates.
Ingestion points: .pawbytes/creative-suites/brands/{brand}/guidelines.md, .pawbytes/creative-suites/brands/{brand}/campaigns/{campaign}/brief.md, and .pawbytes/creative-suites/brands/{brand}/campaigns/{campaign}/scripts/.
Boundary markers: The instructions do not define explicit boundary markers or 'ignore' commands when interpolating these files into downstream prompts or code.
Capability inventory: The skill has access to network operations (curl), file system writes, and dynamic rendering tools (Puppeteer/Playwright).
Sanitization: No explicit sanitization or validation of the ingested text is performed before interpolation into templates or AI prompts.
[DYNAMIC_EXECUTION]: The skill uses Puppeteer or Playwright to render HTML/CSS templates into final images. The rendering process involves dynamic script generation where variables (such as headlines and background image URLs) are injected into HTML source code before being executed in a headless browser environment. The provided example code in template-examples.md uses the --no-sandbox flag, which reduces browser isolation.
[COMMAND_EXECUTION]: The skill relies on the curl command-line tool to interact with the fal.ai API for image generation, job status polling, and asset retrieval.
[EXTERNAL_DOWNLOADS]: The skill fetches generated visual assets from fal.ai infrastructure (queue.fal.run). These operations are consistent with the skill's primary purpose and target well-known AI service endpoints.

paw-cra-agent-designer