The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the claude CLI tool using the --bg and --name flags to spawn a new background agent process from within the current working directory.
[PROMPT_INJECTION]: The skill ingests conversation history and external artifacts such as specs, plans, and architectural decision records (ADRs) to generate a summary that serves as the seed prompt for a new agent. This creates a surface for indirect prompt injection where malicious instructions stored in project files could influence the behavior of the spawned agent.
Ingestion points: The agent is instructed in SKILL.md to process the "current conversation" and external files referenced by path or URL.
Boundary markers: There are no specific instructions to use delimiters or ignore embedded instructions within the summarized handoff content.
Capability inventory: The skill triggers shell execution via the claude CLI.
Sanitization: The instructions include a manual safety requirement for the agent to redact sensitive information such as API keys, passwords, or PII before passing the summary to the new agent.

claude-handoff