The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from an external source (Google Search Console via the GSC Wizard MCP server), including URLs, page names, and search queries. This creates a potential surface for indirect prompt injection if an attacker can manipulate the search data to include instructions for the agent.
Ingestion points: Data enters the agent context through tools such as get_migration_health, compare_migration, and list_migration_redirects (SKILL.md).
Boundary markers: The instructions do not specify the use of delimiters or 'ignore embedded instructions' warnings when handling this external content.
Capability inventory: The skill has the ability to read site data, perform bulk URL inspections, and write operations like creating annotations or managing redirect mappings (SKILL.md).
Sanitization: There are no explicit instructions for sanitizing, escaping, or validating the external content before processing it.
[SAFE]: The skill includes safety-conscious features such as requiring user confirmation before performing destructive actions (like deleting redirect mappings) and monitoring tool quotas to avoid service exhaustion.

site-migration