The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted conversation history and user-supplied purposes, creating an indirect prompt injection surface. Malicious instructions within the conversation could theoretically be included in the generated handoff file.
Ingestion points: $ARGUMENTS and conversation history.
Boundary markers: Employs <purpose> and <analysis> tags for structure.
Capability inventory: Writes markdown files to the local .claude/handoffs/ directory.
Sanitization: Content is summarized and reorganized rather than strictly sanitized.
[COMMAND_EXECUTION]: The skill instructs the agent to perform file system operations, specifically creating a directory and writing markdown files. These actions are limited to the local .claude/handoffs/ path and align with the skill's stated purpose of persisting session state.

handoffs