The Agent Skills Directory

Indirect Prompt Injection (HIGH): The skill is explicitly instructed to ingest and process external, untrusted content from 'competitor' websites and sitemaps (e.g., in references/semantic_topical_map.md and references/seo_audit_protocol.md).
Ingestion points: Uses read_file and potentially run_command (for scraping) to access external data.
Boundary markers: There are no instructions defining clear boundaries or ignoring embedded instructions within the data being audited.
Capability inventory: The agent has write_file, edit_file, and run_command permissions, allowing it to modify the local codebase based on instructions found in audited data.
Sanitization: No evidence of sanitization or validation of the content being read before it is used to guide file edits or command execution.
Command Execution (MEDIUM): The skill uses run_command for SEO tasks like running npx lighthouse (references/seo_audit_checklist.md). While legitimate for the role, this provides an execution vector if the command parameters are influenced by malicious data ingested during an audit.

resonance-seo