The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides scripts such as multi_machine_search.sh and recover.sh that utilize shell commands to perform indexing, search operations, and diagnostics. It specifically uses ssh and rsync to connect to and execute commands on remote hosts to facilitate distributed session searching across multiple machines.
[PROMPT_INJECTION]: The skill's primary function is to process and analyze historical agent session logs (JSONL files). This functionality presents an indirect prompt injection surface, as the logs contain untrusted content from past interactions that could influence the agent's reasoning if malicious instructions were present in the history.
Ingestion points: Agent session history files located in harness-specific directories (e.g., ~/.claude/projects/, ~/.codex/sessions/) are read and processed by cass search, cass view, and cass expand.
Boundary markers: There are no explicit technical boundary markers or 'ignore' instructions implemented when presenting retrieved historical content to the agent.
Capability inventory: The execution environment includes shell access, SSH connectivity for remote search, and file system write access for managing its own index and configuration data.
Sanitization: The skill does not perform explicit sanitization or filtering of the historical content before it is presented to the agent for archaeology or mining.

cass