The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use shell utilities such as shasum, python3, jq, and awk for hashing prompts, validating JSON transcript files, and extracting structured data from logs. These operations are used to maintain data integrity and consistency throughout the debate process.
[PROMPT_INJECTION]: The skill facilitates an indirect prompt injection surface by design, as it ingests data ('deltas') from subagents and projects them into shared and private context files that are included in future turn prompts. The skill mitigates associated risks through isolated subagent environments, explicit structural markers (fenced blocks) for data extraction, and human-in-the-loop checkpoints.

moderate-debate