Back to Details

brain-load

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Brain Load

Brain 加载

Read project context from the Claude Brain Logseq graph and present it to Claude so work can continue seamlessly across sessions and devices.
从Claude Brain Logseq图谱中读取项目上下文并呈现给Claude，以便跨会话和设备无缝继续工作。

Prerequisites

前提条件

Resolve the graph path per 
skills/_shared/path-resolution.md
.
按照
skills/_shared/path-resolution.md
解析图谱路径。

Loading a Specific Project

加载特定项目

When the user says "load <project>" or similar:
  1. Find the project page using the algorithm in 
    references/matching.md
    .
  2. Read the matched project page using the section-targeted-read pattern in 
    skills/_shared/section-locator.md
    . In brief mode, read only Properties + Overview + Current Plan + last 3 Session Log entries (target ≤80 total lines). In full mode, read all sections but still section-by-section, never as a single big Read.
  3. Read related context (full mode only). Follow 
    [[link]]
    references inside the project page if they're critical (e.g., a linked design doc). Skip in brief mode.
  4. Read today's journal (
    journals/yyyy_MM_dd.md
    ) if it exists — captures session notes from earlier today.
  5. Check active Jira tasks. If the project's Current Plan section references task IDs (e.g., 
    PROJ-1234
    ), and the task folder is accessible, optionally read 
    plan.md
    . Full mode or explicit user ask only — in brief mode, just mention task IDs and status.
  6. Apply staleness rules. Use 
    skills/_shared/staleness.md
    against the project's 
    last-updated::
    and 
    status::
    .
  7. Surface session continuity hints. From the most recent Session Log entry: what was last worked on, any 
    open-questions::
    , any blockers. Frame as: "Picking up where you left off — last session ([date]): [summary]. Open questions: [list]."
  8. Present a context summary conversationally: project name + status + staleness, current plan/task, continuity hint, key recent decisions (full mode only), suggested next actions.
  9. Write a journey-log entry per 
    skills/_shared/journey-log.md
    with activity line: 
    loaded [[Projects/<MatchedName>]] (brief)
    or 
    loaded [[Projects/<MatchedName>]] (full)
    .
当用户说"load <project>"或类似表述时：
  1. 查找项目页面：使用
    references/matching.md
    中的算法。
  2. 读取匹配的项目页面：采用
    skills/_shared/section-locator.md
    中的分段读取模式。简洁模式下，仅读取属性（Properties）+ 概述（Overview）+ 当前计划（Current Plan）+ 最近3条会话日志（Session Log）条目（目标≤80行）。完整模式下，读取所有章节，但仍需逐段读取，切勿一次性读取整页内容。
  3. 读取相关上下文（仅完整模式）：如果项目页面中的
    [[link]]
    引用至关重要（例如链接的设计文档），则跟进读取。简洁模式下跳过此步骤。
  4. 读取今日日志：如果
    journals/yyyy_MM_dd.md
    存在，则读取该文件——捕获今日早些时候的会话笔记。
  5. 检查活跃Jira任务：如果项目的当前计划章节引用了任务ID（例如
    PROJ-1234
    ），且任务文件夹可访问，则可选择读取
    plan.md
    。仅在完整模式或用户明确要求时执行——简洁模式下，仅提及任务ID和状态。
  6. 应用过期规则：根据项目的
    last-updated::
    和
    status::
    字段，使用
    skills/_shared/staleness.md
    中的规则。
  7. 显示会话连续性提示：从最新的会话日志条目中提取：上次工作内容、任何
    open-questions::
    、任何阻塞点。表述为："接续上次工作——上次会话（[日期]）：[摘要]。待解决问题：[列表]。"
  8. 以对话形式呈现上下文摘要：项目名称+状态+过期情况、当前计划/任务、连续性提示、关键近期决策（仅完整模式）、建议后续操作。
  9. 写入journey-log条目：按照
    skills/_shared/journey-log.md
    的要求，记录活动行：
    loaded [[Projects/<MatchedName>]] (brief)
    或 
    loaded [[Projects/<MatchedName>]] (full)
    。

Loading General Context

加载通用上下文

When the user says "load brain" without a project:
  1. Read 
    pages/Index.md
    for the list of all projects.
  2. Read 
    pages/Meta.md
    for user preferences and conventions.
  3. Present a high-level overview: active projects, cross-project decisions, user preferences.
  4. Write a journey-log entry: 
    loaded brain overview
    .
当用户未指定项目而说"load brain"时：
  1. 读取
    pages/Index.md
    获取所有项目列表。
  2. 读取
    pages/Meta.md
    获取用户偏好和约定。
  3. 呈现高层概述：活跃项目、跨项目决策、用户偏好。
  4. 写入journey-log条目：
    loaded brain overview
    。

Searching Across the Brain

在Brain中搜索

When the user asks "what do we know about X" or similar, follow the algorithm in 
references/search.md
. After presenting findings, write a journey-log entry: 
searched "X" · N hits
.
当用户询问"what do we know about X"或类似问题时，遵循
references/search.md
中的算法。呈现结果后，写入journey-log条目：
searched "X" · N hits
。

Load Modes: Brief vs Full

加载模式：简洁模式 vs 完整模式

Brief mode (default — "load <project>"):
  1. Properties (type, status, last-updated)
  2. Overview (first 5 bullets only)
  3. Current Plan (full)
  4. Session Log (last 3 entries only)
  5. Skip Implementation, Decisions, linked context
Full mode ("load <project> full", "load everything about <project>"):
  1. All properties
  2. Full Overview
  3. Full Current Plan
  4. Full Implementation
  5. All Decisions
  6. Last 10 Session Log entries
  7. Related context pages (follow 
    [[links]]
    )
  8. Today's journal entry
In brief mode, mention that more is available: "Loaded brief context. Say 'load full' for decisions, implementation details, and full history."
简洁模式（默认——"load <project>"）：
  1. 属性（类型、状态、最后更新时间）
  2. 概述（仅前5个要点）
  3. 当前计划（完整内容）
  4. 会话日志（仅最近3条条目）
  5. 跳过实现（Implementation）、决策（Decisions）、关联上下文
完整模式（"load <project> full"、"load everything about <project>"）：
  1. 所有属性
  2. 完整概述
  3. 完整当前计划
  4. 完整实现内容
  5. 所有决策
  6. 最近10条会话日志条目
  7. 关联上下文页面（跟进
    [[links]]
    ）
  8. 今日日志条目
简洁模式下，需提及可获取更多内容："已加载简洁上下文。如需查看决策、实现细节和完整历史，请说'load full'。"

Important Notes

重要注意事项

  • Be selective about reads. Token budget matters — use the section-targeted-read pattern in step 2 of "Loading a Specific Project" rather than reading whole pages.
  • After loading, confirm what's available and suggest next actions.
  • All graph content is Logseq outliner format. See 
    CLAUDE.md
    for invariants.
  • 读取内容时需有所选择。Token预算至关重要——采用“加载特定项目”步骤2中的分段读取模式，而非读取整页内容。
  • 加载完成后，确认可用内容并建议后续操作。
  • 所有图谱内容均为Logseq大纲格式。请查看
    CLAUDE.md
    了解不变规则。