To ensure the integrity of the agent's memory core, this ticket covers updating the .github/WORKING_WITH_AGENTS.md guide to instruct human developers on their role in the memory-saving process.
While the agent is mandated to follow a "save-then-respond" protocol, a sufficiently derailed agent might fail to do so. This update introduces the concept of a "human-in-the-loop" safeguard for the memory protocol itself.
Acceptance Criteria
- A new section, titled "The Memory Core: A Shared Responsibility", is added to
.github/WORKING_WITH_AGENTS.md.
- This section explains the agent's "save-then-respond" duty.
- This section explicitly defines the human developer's duty to verify that the
ai:add-memory tool call is successfully made after every agent turn when the memory core is active.
- The section provides a clear recovery prompt for the user to issue in case the agent fails to save its memory.
To ensure the integrity of the agent's memory core, this ticket covers updating the
.github/WORKING_WITH_AGENTS.mdguide to instruct human developers on their role in the memory-saving process.While the agent is mandated to follow a "save-then-respond" protocol, a sufficiently derailed agent might fail to do so. This update introduces the concept of a "human-in-the-loop" safeguard for the memory protocol itself.
Acceptance Criteria
.github/WORKING_WITH_AGENTS.md.ai:add-memorytool call is successfully made after every agent turn when the memory core is active.