The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's documentation provides a series of shell scripts intended for manual use by the human mediator to verify the integrity and completeness of the review logs. These scripts utilize standard system utilities such as grep, test, and wc to check for the presence of required synthesis blocks, H/M/L findings, and mediator approvals. They do not perform any destructive or privileged operations.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external artifacts (e.g., code diffs, plans, specifications) as part of the review process. The instructions mitigate the risk of indirect prompt injection by explicitly classifying external expert feedback as 'signal to be evaluated' rather than direct instruction. It enforces a strict security boundary by requiring a human 'Mediator' to approve all synthesis artifacts and design decisions, ensuring that no external input can be autonomously executed or adopted without oversight.

consult-outside-expert