The Agent Skills Directory

[SAFE]: The skill's functionality is limited to reading and summarizing local session history from files located in the .maestro/sessions/ and .maestro/decisions.jsonl directories. No dangerous tools or actions are involved.
[DATA_EXFILTRATION]: The skill uses local file access to generate its summaries. There are no network requests, external API calls, or patterns indicating that session data is being exfiltrated to a third party.
[PROMPT_INJECTION]: The directives to follow the 'Context Gathering Protocol' and 'Invoke /agent-workflow' are standard internal workflow instructions. They do not contain markers associated with bypassing safety guidelines or overriding the agent's core instructions.
[COMMAND_EXECUTION]: The skill does not invoke shell commands, execute scripts, or perform any dynamic code evaluation. It relies solely on the agent's ability to read text and provide a formatted summary.

recap