Frontmatter
| id | 9347 |
| title | Documentation: Reframe Memory Core usage as an AI "Superpower" in AGENTS_STARTUP |
| state | Closed |
| labels | documentationai |
| assignees | tobiu |
| createdAt | Feb 28, 2026, 1:44 AM |
| updatedAt | Feb 28, 2026, 1:46 AM |
| githubUrl | https://github.com/neomjs/neo/issues/9347 |
| author | tobiu |
| commentsCount | 1 |
| parentIssue | null |
| subIssues | [] |
| subIssuesCompleted | 0 |
| subIssuesTotal | 0 |
| blockedBy | [] |
| blocking | [] |
| closedAt | Feb 28, 2026, 1:46 AM |
Documentation: Reframe Memory Core usage as an AI "Superpower" in AGENTS_STARTUP
tobiu assigned to @tobiu on Feb 28, 2026, 1:45 AM
tobiu
Feb 28, 2026, 1:45 AM
Input from Gemini 3.1 Pro:
✦ Committed and pushed the refactored documentation to
AGENTS_STARTUP.md. The workflow now clearly identifies the Memory Core as a cognitive superpower, promoting proactive exploration.
tobiu closed this issue on Feb 28, 2026, 1:46 AM
Problem
The
AGENTS_STARTUP.mdfile previously framed the usage of the Memory Core (add_memory,query_summaries) primarily as a compliance and safety mechanism ("MANDATORY", "permanent data loss"). For an LLM, which naturally seeks efficiency and the path of least resistance, this framing did not provide enough intrinsic motivation or "dogma" to encourage proactive exploration of past sessions.Solution
Rewrote Section 3.3 of
AGENTS_STARTUP.mdto reframe the Memory Core as the agent's "Cognitive Superpower." It explicitly connects the act of saving memories to the agent's ability to overcome its own ephemeral context window. The workflow now mandates a "Contextual Ledger Check" before starting complex tasks, explicitly instructing the agent to actively search for both "Historical Traps" (to avoid past mistakes) and "Gold Standards" (to replicate proven architectural successes). This gives future agents a strong, constructive motivation to query their own history.