The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted conversation history and user arguments to generate instructions for a future agent session. -- Ingestion points: Current conversation context and user-provided arguments as described in SKILL.md. -- Boundary markers: The instructions do not specify the use of delimiters or 'ignore embedded instructions' warnings in the generated handoff document to separate untrusted summary content from agent instructions. -- Capability inventory: The agent is instructed to write the resulting document to the operating system's temporary directory, involving file-system write operations. -- Sanitization: The skill includes a proactive safety instruction to redact sensitive data like API keys and passwords, but lacks mechanisms to filter or escape malicious instructions that may have been injected into the conversation history.

handoff