The Agent Skills Directory

[PROMPT_INJECTION]: The skill is subject to indirect prompt injection risks as its primary function involves processing content from untrusted external sources (websites, GitHub repositories, PDFs, and video transcripts).
Ingestion points: External data enters the agent context through tools such as scrape_docs, scrape_github, scrape_pdf, scrape_video, and scrape_codebase (SKILL.md).
Boundary markers: The skill definition does not include specific delimiters or instructions for the agent to ignore potentially malicious instructions embedded in the scraped content.
Capability inventory: The skill facilitates network operations (scrape_docs, upload_skill), local file access (scrape_codebase), and system workflows (install_skill, package_skill).
Sanitization: No explicit sanitization or validation of the retrieved content is defined in the skill instructions.

skill-builder