The Agent Skills Directory

Indirect Prompt Injection: The skill processes live device events and error logs from an external iOS application via local debugging tools. This creates an interface where the agent might process untrusted data.
Ingestion points: The agent receives external data through tools such as wait_for_events, get_errors, and get_sdk_state as described in the Normal agent loop in SKILL.md.
Capability inventory: The agent has the ability to connect to local servers using connect_to_debug_server and can aggregate data via export_diagnostic_bundle in SKILL.md.
Boundary markers: While the skill provides clear operational instructions, it does not specify explicit delimiters or "ignore embedded instructions" warnings for the external data it ingests.
Sanitization: The instructions proactively suggest collecting a "redacted bundle" for handoff, which is a positive measure to reduce sensitive data exposure during the diagnostic process.

live-debugging-mcp