The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements a workflow that ingests data from external, untrusted sources which represents a potential attack surface.
Ingestion points: The skill uses WebFetch in Step 3 and Step 6 to extract data from various external platforms including business websites, Yelp, Facebook, and competitor profiles.
Boundary markers: The instructions do not define explicit boundary markers or delimiters (like XML tags or clear 'ignore' instructions) when processing the content fetched from external URLs.
Capability inventory: The skill configuration in the frontmatter allows the use of Bash (command execution), WebFetch (network operations), and Read (filesystem access).
Sanitization: While the instructions direct the agent to extract specific fields (NAP data, categories, and reviews), there is no explicit instruction to sanitize or filter the content for embedded prompt instructions or malicious code before processing.
[EXTERNAL_DOWNLOADS]: The skill is designed to perform network operations to fetch configuration and business data.
The process involves accessing official Google Business Profile guidelines and various well-known directory services (Yelp, Apple Maps, Bing, Facebook) to perform Name, Address, and Phone (NAP) consistency audits.
The skill also involves fetching and analyzing data from competitor websites to perform benchmarking.

30x-seo-local