The Agent Skills Directory

[PROMPT_INJECTION]: The skill describes a documentation workflow that ingests data from external websites (Phase 1) and uses it to generate content and code (Phase 2). This creates a surface for indirect prompt injection where malicious instructions embedded in third-party articles or videos could influence the agent's behavior. • Ingestion points: External URLs curated for MDN, articles, and videos in SKILL.md. • Boundary markers: The skill does not define specific delimiters or instructions to ignore embedded commands in the fetched research content. • Capability inventory: The agent has capabilities to write files (/docs/concepts/) and execute shell commands (npm test). • Sanitization: No specific input sanitization or filtering logic is provided to handle content fetched from external sources.
[COMMAND_EXECUTION]: The workflow directs the agent to run npm test (Phase 3 and 4) to verify code examples. While this is an intended quality gate for documentation, it involves executing code derived from external research data.
[EXTERNAL_DOWNLOADS]: The skill instructs the agent to visit and verify external resources, including well-known documentation services like MDN and various third-party article and video platforms, as part of the research phase.

concept-workflow