The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and summarize data from untrusted sources, creating a risk that malicious instructions embedded in websites could influence downstream agent behavior.
Ingestion points: The skill collects data through user-provided answers in Phase 1 and performs automated web research using the web_search and web_fetch tools in Phase 2 to retrieve content from external domains.
Boundary markers: There are no instructions for the agent to use delimiters or explicit "ignore embedded instructions" markers when processing the retrieved web content.
Capability inventory: The skill utilizes web retrieval tools (web_search, web_fetch) and filesystem operations (saving a markdown file and using present_files) to persist and share the generated context.
Sanitization: The skill lacks mechanisms to sanitize or validate the content retrieved from external websites before incorporating it into the brand-context.md file, which is intended to be a trusted source for other modules.

brand-intake