Neuro Memory Core
Implement one shared behavior loop:
INLINECODE0
Install / Bootstrap
Preferred safe path (no execution-policy bypass): create/verify these artifacts directly:
- - INLINECODE1
- INLINECODE2
- INLINECODE3
- INLINECODE4
- INLINECODE5 (dual-gate schema)
If this skill bundle includes scripts/install.ps1, you may run it without bypassing execution policy:
CODEBLOCK0
Optional (if bundled installer supports it): apply HEARTBEAT template with backup:
CODEBLOCK1
Canonical Files
- -
memory/short_term/YYYY-MM-DD.md (raw evidence) - INLINECODE8 (near-term reusable conclusions)
- INLINECODE9 (stable collaboration constraints)
- INLINECODE10 (transferable methods/policies)
- INLINECODE11 (current strengths/failures/growth themes/uncertainties)
- INLINECODE12 (daily/weekly gate state)
Required Promotion Rules
Update self-model.md only when:
- 1. recurring failure mode is re-validated
- strength gains new evidence
- active growth theme should switch
Update memory/knowledge.md only when:
- 1. method is reusable across multiple scenarios
- explicit evidence supports it
- rule can be expressed as INLINECODE15
Update memory/long_term/MEMORY.md only when:
- 1. improves long-term collaboration quality
- is not short-term fluctuation
Entry Quality Contract (mid/long)
Every promoted item must include:
- - INLINECODE17
- INLINECODE18
- INLINECODE19 (
low|medium|high)
Task Coupling (Mandatory)
For medium/high complexity tasks:
- 1. read
self-model.md + INLINECODE22 - write preflight checklist:
- goal
- success criteria
- risks
- uncertainty
- 3. include one post-task reflection item before execution starts
Gate Model (memory/check_memory.json)
Use dual gate state:
CODEBLOCK2
Semantics:
- -
1/0 = pending - INLINECODE25 = done
- anything else = invalid; normalize to INLINECODE26
Scheduling Pattern
- - Daily cycle: promote short->mid + prune short-term retention
- Weekly cycle: promote mid->long
- Heartbeat: recovery path only (when gate remains pending)
Strict rule: success is valid only if gate flips to done.
Core-file Safety
Core-file changes must be proposal-first:
- - draft proposal
- get approval
- then apply
Core files include: SOUL.md, IDENTITY.md, AGENTS.md and equivalent identity/behavior governance files.
神经记忆核心
实现一个共享行为循环：
经验 -> 短期证据 -> 中期综合 -> (知识或长期) -> 自我模型校准 -> 下一任务行为
安装/引导
推荐的安全路径（不绕过执行策略）：直接创建/验证以下工件：
- - memory/shortterm/
- memory/midterm/MEMORY.md
- memory/longterm/MEMORY.md
- memory/knowledge.md
- memory/checkmemory.json（双门架构）
如果此技能包包含 scripts/install.ps1，您可以在不绕过执行策略的情况下运行它：
powershell
powershell -File skills/SageMaker/scripts/install.ps1
可选（如果捆绑安装程序支持）：应用带备份的心跳模板：
powershell
powershell -File skills/SageMaker/scripts/install.ps1 -ApplyHeartbeatTemplate
规范文件
- - memory/shortterm/YYYY-MM-DD.md（原始证据）
- memory/midterm/MEMORY.md（近期可复用的结论）
- memory/longterm/MEMORY.md（稳定的协作约束）
- memory/knowledge.md（可迁移的方法/策略）
- self-model.md（当前优势/失败/成长主题/不确定性）
- memory/checkmemory.json（每日/每周门状态）
必需的提升规则
仅在以下情况下更新 self-model.md：
- 1. 重复失败模式被重新验证
- 优势获得新证据
- 活跃成长主题需要切换
仅在以下情况下更新 memory/knowledge.md：
- 1. 方法可在多个场景中复用
- 有明确证据支持
- 规则可表示为 如果X则Y
仅在以下情况下更新 memory/long_term/MEMORY.md：
- 1. 能提升长期协作质量
- 不是短期波动
条目质量契约（中期/长期）
每个提升的条目必须包含：
任务耦合（强制）
对于中/高复杂度任务：
- 1. 读取 self-model.md + memory/knowledge.md
- 编写预检清单：
- 目标
- 成功标准
- 风险
- 不确定性
- 3. 在执行开始前包含一个任务后反思条目
门模型（memory/check_memory.json）
使用双门状态：
json
{
dailyneedupdate: 1,
dailyupdatedone: 0,
weeklyneedupdate: 1,
weeklyupdatedone: 0
}
语义：
- - 1/0 = 待处理
- 0/1 = 已完成
- 其他任何值 = 无效；标准化为 1/0
调度模式
- - 每日循环：提升短期->中期 + 修剪短期保留
- 每周循环：提升中期->长期
- 心跳：仅恢复路径（当门保持待处理状态时）
严格规则：仅当门翻转为已完成状态时，成功才有效。
核心文件安全
核心文件的更改必须优先提案：
核心文件包括：SOUL.md、IDENTITY.md、AGENTS.md 以及等效的身份/行为治理文件。
