The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's primary function is to construct and execute a shell command to spawn background processes using the system's CLI.
Evidence: The instruction to execute claude --bg --name "<descriptive name>" "<handoff summary>" in SKILL.md.
[INDIRECT_PROMPT_INJECTION]: The skill ingests the entire conversation history and user-provided arguments to generate a 'handoff summary' which serves as the system prompt for a new agent.
Ingestion points: The current conversation history and user-supplied handoff arguments (SKILL.md).
Boundary markers: None. The summary is passed directly as a string argument to the next agent.
Capability inventory: The skill can execute shell commands via the claude CLI.
Sanitization: The instructions recommend redacting sensitive data like API keys, but there are no technical controls to prevent prompt injection from being carried over into the new agent's context.
[DATA_EXFILTRATION]: Passing potentially long conversation summaries as command-line arguments can lead to data exposure, as command arguments are often logged by the operating system or visible to other users via process monitoring tools like ps or top.
Evidence: The use of the summary string inside the command execution template in SKILL.md.

claude-handoff