The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data in the form of code diffs, branch descriptions, and pull request content to generate explanations and quizzes.
Ingestion points: Data enters the context via the code diffs and PR metadata provided to the agent for summarization in SKILL.md.
Boundary markers: The instructions lack explicit delimiters or instructions to ignore embedded prompts within the diff content, which could allow an attacker to influence the documentation style or quiz content through comments in the code.
Capability inventory: The skill has the capability to execute a local Python script (render.py) to generate HTML files and can write to the filesystem (e.g., /tmp).
Sanitization: The render.py script utilizes html.escape() to sanitize text content before embedding it into the HTML output, which mitigates basic cross-site scripting (XSS) risks in the generated reports.
[EXTERNAL_DOWNLOADS]: The render.py script generates HTML that references external resources from a well-known service.
Evidence: The script includes hardcoded URLs for highlight.js CSS and JS files hosted on cdnjs.cloudflare.com.
Context: These downloads are from a recognized, official CDN and are standard for providing syntax highlighting in web-based documentation.

explain-diff