The Agent Skills Directory

[SAFE]: The skill is purely instructional and focuses on information architecture. No malicious patterns, obfuscation, or security vulnerabilities were detected across any of the files.
[PROMPT_INJECTION]: The instructions establish a professional persona as an information architecture expert. There are no attempts to override safety filters, disregard previous instructions, or implement 'jailbreak' patterns.
[DATA_EXFILTRATION]: The skill does not contain any network operations (e.g., curl, fetch) or access sensitive file paths (e.g., .env, .ssh). It reads from standard agent context files like .agents/product-marketing-context.md for functional purposes.
[COMMAND_EXECUTION]: No shell commands, subprocess calls, or privilege escalation attempts (e.g., sudo) were found in the skill's instructions or templates.
[REMOTE_CODE_EXECUTION]: The skill does not download or execute external scripts. It relies on internal markdown templates and Mermaid diagrams for its functionality.
[INDIRECT_PROMPT_INJECTION]: While the skill ingests data from local context files, it lacks dangerous capabilities (like network access or command execution) that would make it a significant target for indirect injection attacks.

site-architecture