The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied requests to generate system architectures and implementation plans. A malicious request could attempt to influence the design outputs (e.g., suggesting insecure sync policies or conflict resolution). The risk is low as the output is primarily documentation and planning text intended for developer review.
Ingestion points: Processes user-provided requirements in SKILL.md to create architecture summaries and project contracts.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands in input data are defined.
Capability inventory: Generates architecture summaries, project contracts, scaffold layouts, and implementation plans. Instructions do not mandate the use of shell execution or network tools for design tasks.
Sanitization: No input validation or sanitization of the user's initial design request is implemented.
[EXTERNAL_DOWNLOADS]: The skill's documentation and verification scripts reference Node.js packages and external rulesets. These are used for development and maintenance.
Evidence: README.md and CONTRIBUTING.md suggest installing markdownlint-cli and compose-agentsmd. agent-ruleset.json and AGENTS.md reference rules sourced from the vendor's GitHub repository (github:metyatech/agent-rules). These are recognized as vendor-owned resources.

human-ai-system-builder