The Agent Skills Directory

[SAFE]: The skill provides structured instructions and templates for information architecture and site planning. It does not contain executable code, remote downloads, or dangerous command execution patterns.
[DATA_EXPOSURE]: The instructions direct the agent to check for local configuration files (e.g., .agents/product-marketing.md) to gather project context. This is a standard practice for context-aware developer tools and does not involve unauthorized access or exfiltration of sensitive credentials.
[INDIRECT_PROMPT_INJECTION]: The skill defines a surface for processing external data by reading project-specific marketing files. While these files are external to the skill itself, the risk is negligible as the skill lacks high-privilege capabilities such as network access, file writing, or shell execution.
Ingestion points: .agents/product-marketing.md, .claude/product-marketing.md in SKILL.md.
Boundary markers: Not explicitly defined for these files.
Capability inventory: None. The skill only generates architectural plans and diagrams in text/Mermaid format.
Sanitization: Not applicable as no code is executed.

site-architecture