The Agent Skills Directory

[PROMPT_INJECTION]: No malicious override or bypass patterns detected. The instructions follow standard expert-persona roleplay for information architecture.
[DATA_EXFILTRATION]: No network operations or hardcoded credentials detected. The skill only requests reading local marketing context files to personalize its output.
[REMOTE_CODE_EXECUTION]: No remote script downloads or piped command executions are present. The skill does not install any packages.
[COMMAND_EXECUTION]: No subprocess spawning, shell commands, or dynamic execution patterns found. The skill's output is limited to text, tables, and Mermaid diagrams.
[OBFUSCATION]: Content is entirely in plain text with no encoded strings (Base64, hex) or hidden characters detected.
[INDIRECT_PROMPT_INJECTION]: While the skill reads external context from .agents/product-marketing.md, it lacks any capabilities (like file writing or network access) that could be exploited via malicious content in those files.

site-architecture