The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from Google Search Console, including search queries, page titles, and site annotations. While this constitutes an attack surface, the risk is limited to the agent's interpretation of SEO data.
Ingestion points: The tools query_top_queries, query_top_pages, and list_annotations fetch external data.
Boundary markers: No explicit delimiters are defined in the instructions to separate fetched data from agent instructions.
Capability inventory: Capabilities include network-based report sharing (create_shared_report) and formatted HTML generation (generate_seo_report).
Sanitization: The instructions do not specify sanitization or escaping of the fetched metrics or strings before they are used in report generation.
[DATA_EXFILTRATION]: The skill provides tools for sharing SEO digests via external reports. This is an intended feature of the skill and is documented with appropriate safety context. The instructions mandate a multi-step process for granting access (create_report_client and manage_report_access) and explicitly warn the user that reports are private by default, which serves as a safeguard against accidental data exposure.

weekly-report