The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for executing the claude CLI command to spawn background agents. It employs shell command substitution $(cat <summary file>) to pass the contents of a generated summary file as a single argument to the command-line interface.
[INDIRECT_PROMPT_INJECTION]: The skill ingests the current conversation history and user-supplied arguments to generate a prompt for a new background agent. This creates a potential surface where malicious or confusing instructions within the conversation could be persisted into the new agent's context.
Ingestion points: The skill reads the full current conversation history and any specific user arguments provided at invocation as stated in SKILL.md.
Boundary markers: No specific delimiters or boundary markers are required for the output summary; however, the skill uses a file-passing mechanism (-- "$(cat file)") to prevent the shell from interpreting characters within the data.
Capability inventory: The skill enables the creation and execution of background agent processes via the claude binary.
Sanitization: The instructions explicitly mandate the redaction of sensitive information, such as API keys, passwords, and personally identifiable information (PII), before the summary is used to seed the new agent.

claude-handoff