The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from external API tools and user-provided sources, which represents a potential vector for indirect instructions.
Ingestion points: Data retrieved through the sandbase_run and sandbase_run_get tools, along with source material identified in the review workflow.
Boundary markers: There are no explicit instructions for the agent to utilize boundary markers or specific "ignore" directives when handling data from external capabilities.
Capability inventory: The skill utilizes a suite of tools including sandbase_discover, sandbase_inspect, sandbase_run, sandbase_run_get, sandbase_runs, and sandbase_account to interact with external services and manage implementation work.
Sanitization: The instructions do not define specific sanitization, filtering, or validation steps for content returned by external tools prior to its incorporation into the agent's context.

accessibility-review