The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the ralph CLI and standard shell utilities (jq, grep, sed, ls) to manage loop states, inspect logs, and process diagnostic data.
[INDIRECT_PROMPT_INJECTION]: The skill processes data generated by other agent loops, which creates an attack surface for indirect prompt injection.
Ingestion points: The skill reads loop logs, history, diffs, and diagnostic artifacts such as agent-output.jsonl and prompt-log.md (documented in SKILL.md and references/diagnostics.md).
Boundary markers: No explicit delimiters or instructions to ignore embedded commands within the processed logs are specified.
Capability inventory: The skill can execute various ralph CLI commands (run, merge, resume, retry) based on the data it inspects.
Sanitization: The instructions do not specify any sanitization or validation of the content retrieved from loop logs or diagnostic files before it is processed by the agent.

ralph-loop