The Agent Skills Directory

[DYNAMIC_EXECUTION]: The build sub-skill contains a React component example ('CodeSandbox.tsx') in 'skills/tutorial-writer-build/SKILL.md' that uses 'eval(code)' to execute arbitrary JavaScript. While the documentation includes a warning that a secure sandbox should be used in production, an agent following this template might inadvertently introduce a critical remote code execution vulnerability into the user's project.
[INDIRECT_PROMPT_INJECTION]: The skill architecture relies on fetching external data from the web during the 'research' phase to inform the 'writing' phase.
Ingestion points: External technical data is gathered via 'WebSearch' and stored in 'reference/' directories (defined in 'skills/tutorial-writer-research/SKILL.md').
Boundary markers: The skill does not define specific delimiters or instructions to ignore embedded prompts within the gathered research materials.
Capability inventory: The skill has the capability to write files to the local directory ('book/', 'web/'), execute shell commands ('npm run build'), and configure GitHub Actions for deployment.
Sanitization: There is no explicit mention of sanitizing or escaping the content fetched from the web before it is processed by the agent to generate tutorial chapters.
[COMMAND_EXECUTION]: The skill frequently uses shell commands for project management and deployment in 'skills/tutorial-writer-build/SKILL.md' and 'skills/tutorial-writer-publish/SKILL.md', including 'npm install', 'npm run build', and 'npx serve'. While these are standard for web development, they execute in the agent's environment.
[EXTERNAL_DOWNLOADS]: The deployment workflow in 'skills/tutorial-writer-publish/SKILL.md' references the 'withastro/action@v3' GitHub Action. This is a well-known service provided by the official Astro team and is used for standard CI/CD deployment to GitHub Pages.

tutorial-writer