The Agent Skills Directory

[COMMAND_EXECUTION]: The skill directs the agent to use the local claude CLI with the --bg flag to create a new session based on the provided summary.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface by converting conversation history into a system prompt for a subsequent agent instance. Evidence includes: 1. Ingestion point: The current conversation history is the source for the handoff summary. 2. Boundary markers: No delimiters are specified for the untrusted conversation content. 3. Capability inventory: The skill can trigger background process creation via the claude command. 4. Sanitization: The instructions mandate redacting credentials and PII from the output.

claude-handoff