The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources which may contain malicious instructions.
Ingestion points: Uses WebFetch to retrieve content from external URLs and Read to access structured data files (CSV, JSON).
Boundary markers: The instructions lack markers or specific directives for the agent to distinguish between the skill's instructions and the content being analyzed.
Capability inventory: The skill is granted access to the Bash tool for shell execution and the Write tool for file modifications, creating a risk if instructions are injected via the fetched data.
Sanitization: There are no instructions for validating, escaping, or filtering retrieved content before it is processed by the agent.

seo-programmatic