The Agent Skills Directory

[SAFE]: The skill is purely text-based and does not execute scripts, commands, or access external network resources. All logic is contained within the prompt instructions and a local reference file.- [SAFE]: The skill references a local file 'references/experiments.md' which contains static marketing information and no executable elements.- [PROMPT_INJECTION]: The skill ingests untrusted marketing page content for analysis, which presents a surface for indirect prompt injection.
Ingestion points: User-provided text content from marketing pages (homepages, landing pages, blog posts) via the chat context.- Boundary markers: The prompt lacks explicit delimiters or 'ignore' instructions to distinguish the data being analyzed from the system instructions.- Capability inventory: The skill only generates text-based recommendations; it has no tools for network access, filesystem modification, or command execution.- Sanitization: No sanitization or validation of input text is performed.

page-cro