The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an inherent surface for indirect prompt injection due to its requirement to ingest and analyze untrusted data.
Ingestion points: External website content is retrieved using web_fetch and curl commands. Additionally, local context is read from files such as .agents/product-marketing.md.
Boundary markers: The instructions do not specify the use of delimiters (e.g., XML tags or triple quotes) or explicit system-level warnings to the agent to disregard instructions embedded within the analyzed content.
Capability inventory: The skill leverages web retrieval tools (web_fetch, curl) to perform automated site audits.
Sanitization: No procedures for sanitizing, escaping, or validating the ingested content are mentioned in the audit instructions.

seo-audit