The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references numerous external academic papers and technical documentation from trusted sources including Anthropic, OpenAI, Microsoft, and arXiv to support its design recommendations.
References include links to official engineering blogs (anthropic.com, openai.com) and research repositories (arxiv.org, openreview.net).
Includes links to industry-standard guides like the Prompting Guide and agentic-patterns.com.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest user-provided requirements for complex agentic systems and process them into a structured design document.
Ingestion points: User responses to the 12-stage Q&A flow defined in SKILL.md.
Boundary markers: Explicit boundary identification is performed in Stage 3 of the flow to define system constraints.
Capability inventory: The skill only generates markdown text (design documents) and does not perform file writes, network requests, or command execution.
Sanitization: The skill follows a strict internal logic and rubric-based evaluation to validate user inputs against design patterns.
[SAFE]: The skill is primarily instructional and documentation-focused.
Stated purpose as a design partner matches the actual behavior of the provided files.
Code examples provided in the examples/ directory are for demonstration purposes and are not executed by the agent.
No obfuscation, persistence mechanisms, or credential handling was found.

agentic-system-design