The Agent Skills Directory

[COMMAND_EXECUTION]: The skill operates by executing the ralph CLI for starting, stopping, and merging process loops. It also utilizes common system utilities such as ls, head, and jq to parse diagnostic logs and session information.
[PROMPT_INJECTION]: The skill processes log files and JSONL data (e.g., agent-output.jsonl, orchestration.jsonl) which record interactions from the Ralph loop processes. This represents an indirect prompt injection surface as the data might originate from external sources being processed by the loops.
Ingestion points: Accesses log data via ralph loops logs and diagnostic files under .ralph/diagnostics/.
Boundary markers: No explicit delimiters or instructions to ignore embedded content are used when reading these files.
Capability inventory: Includes the ability to execute new runs (ralph run), merge code changes (ralph loops merge), and modify queue states.
Sanitization: No specific sanitization or validation of the log content is described before the data is presented to the agent.

ralph-loop