メニュー
Unified memory skill: immediate mode (/remember) persists knowledge via PKB MCP; maintenance mode (/sleep, GHA cron) runs periodic consolidation — transcript mining, knowledge synthesis, data quality, brain sync.
デフォルトでは、最初にソースを確認する Prompt が選択されています。直接コマンドに切り替えるか、ローカルコピーをダウンロードすることもできます。
インストールを決める前に、SKILL.md と SkillsMP に表示されている付属ファイルをお読みください。
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
直接コマンドでは確認用 Prompt が省略されます。実行前にソースを確認してください。
npx skills add https://github.com/nicsuzor/aops --skill rememberコマンドは1行のまま表示されます。コピー前に横へスクロールして全体を確認してください。
ローカルで確認しますか？SkillsMP が現在取得できるファイルをダウンロードできます。
SOC 職業分類に基づく
Core academicOps skill — institutional memory, strategic coordination, workflow routing, and framework governance. Merges butler (chief-of-staff) with framework development conventions.
Daily note lifecycle — compose and maintain a factual daily note. Reports the state of the day; does not prioritise or recommend. SSoT for daily note structure.
Emergency session bail — fast resume task + short handover, no commit/PR/reflection. For when you (or the user) need a clean context now. Use /end-session for canonical close.
SKILL.md を表示中
| name | remember |
| type | skill |
| category | instruction |
| description | Unified memory skill: immediate mode (/remember) persists knowledge via PKB MCP; maintenance mode (/sleep, GHA cron) runs periodic consolidation — transcript mining, knowledge synthesis, data quality, brain sync. |
| triggers | ["remember this","save to memory","store knowledge","sleep cycle","consolidation","brain maintenance"] |
| modifies_files | true |
| needs_task | false |
| mode | execution |
| domain | ["operations"] |
| allowed-tools | mcp_pkb_create,mcp_pkb_create_memory,mcp_pkb_append,mcp_pkb_get_document,mcp_pkb_search,mcp_pkb_update_task,Bash,Read,Write,Grep,Glob,Edit,Skill,mcp_pkb_pkb_orphans,mcp_pkb_list_tasks,mcp_pkb_graph_stats,mcp_pkb_get_stats,mcp_pkb_get_network_metrics,mcp_pkb_get_task,mcp_pkb_task_search,mcp_pkb_pkb_context,mcp_pkb_batch_reparent,mcp_pkb_find_duplicates,mcp_pkb_batch_merge,mcp_pkb_merge_node,mcp_pkb_complete_task,mcp_pkb_batch_reclassify,mcp_pkb_batch_archive,mcp_pkb_batch_update,mcp_pkb_create_task,mcp_pkb_update_memory,mcp_omcp_messages_search,mcp_omcp_messages_query,mcp_omcp_calendar_list_events |
| owner | pauli |
| version | 4.0.0 |
Provides domain instructions for knowledge capture, persistence, and maintenance in the Personal Knowledge Base (PKB) ($ACA_DATA).
/remember)Write or Edit on files under $ACA_DATA/. Use the PKB MCP toolset.mcp_pkb_search for semantic discovery. Do not use Glob or Grep on $ACA_DATA/.mcp_pkb_create: Create a document (projects, contexts, meeting notes).mcp_pkb_create_memory: Add a memory/note.mcp_pkb_append: Extend an existing document.mcp_pkb_get_document: Read a document.mcp_pkb_search: Search before creating new content — and before asserting a fact. Don't assume; look it up./sleep loop.| Content | Write Via | Target Location |
|---|---|---|
| Epics/projects | mcp_pkb_create(type="epic"|"project") | projects/ |
| Tasks/issues | gh issue create (GitHub primary) | Synced automatically |
| Durable knowledge | mcp_pkb_create or mcp_pkb_create_memory | knowledge/, context/ |
| Session findings | mcp_pkb_update_task on parent task | Task body |
Invoke /framework skill — do NOT write to brain/context/ |
skills/ via framework workflow |
| Content | Location | Notes |
|---|---|---|
| Project metadata | projects/<name>.md | Hub files. Hub-doc type is epic. |
| Project details | projects/<name>/ | Subdirectory. |
| Goals | goals/ | Strategic objectives. |
| Context (about user) | context/ | User preferences, history. |
| Sessions/daily | sessions/ | Daily notes only, type: daily-note. |
| General knowledge | knowledge/<topic>/ | Durable facts (non-user). |
| Meeting/call notes | knowledge/<topic>/ or projects/ | type: meeting-note. |
| Maps of Content (MOCs) | knowledge/ or topic subdirs | Navigational hubs, type: moc. |
$ACA_DATA/.$ACA_DATA/ with correct type frontmatter.Maintain exactly one canonical note per first-class topic (tools, projects, skills, agents, concepts).
mcp_pkb_append in the relevant section.Overview, Usage, Known Issues) only if the topic does not exist.During updates, search for duplicate or peer notes. Merge unique content, retire weaker versions, and update wikilinks. For how to retire — delete the superseded file and trust git history; do not leave superseded_by: pointers — follow the single canonical rule in [[consolidation-procedure]] Step 3.
mcp_pkb_search(query="topic").mcp_pkb_create(
title="Descriptive Title",
body="Content with [[wikilinks]].",
type="note" | "project" | "epic" | "knowledge" | "moc" | "meeting-note",
tags=["tag1", "tag2"]
)
Or for lightweight facts:
mcp_pkb_create_memory(
title="descriptive title",
body="content",
tags=["tag1"]
)
[org/repo#123](url)).## Relationships section:
## Relationships
- [related] [[task-id]] — description
- [upstream-bug] [org/repo#123](url)
For framework tasks, capture generalizable patterns (e.g., "Configuration should be set once at initialization") rather than local implementation details.
Factual claims in episodic notes must be formatted as Obsidian callouts:
> [!observation] Factual statement
> Source: [[source-note]]
> Confidence: <numeric value 0.0 to 1.0>
established (>= 0.8): verified, multiple sources.provisional (0.4 - 0.79): limited evidence.speculative (< 0.4): inference.type: knowledge)Include these frontmatter fields:
sources: (YAML list)synthesized: (ISO date)last_reviewed: (ISO date)confidence: (numeric 0.0 - 1.0)maturity: seedling | budding | evergreenUse MOCs to index and curate links to related notes when a topic area grows to 5+ notes.
title: "MOC: Topic Name"
type: moc
tags: [moc, topic-area]
created: YYYY-MM-DD
last_reviewed: YYYY-MM-DD
To capture memories asynchronously:
Task(
subagent_type="general-purpose", model="haiku",
run_in_background=true,
description="Remember: [summary]",
prompt="Invoke activate_skill(name='remember') to persist: [content]"
)
State the PKB write details directly and concisely:
/sleep)Runs periodic offline consolidation on a GHA workflow or manual trigger.
consolidated: YYYY-MM-DD in frontmatter).First match determines mode:
SLEEP_MODE overrides default (short-loop | full-session).LOOP_INTERVAL_MINUTES <= 30 -> short-loop.short-loop.full-session.See [[references/maintenance-phases]] for detailed sub-agent rules, pacing, and exit protocols.
| Phase | Name |
|---|---|
| 0 | Graph Health Baseline |
| 1 | Session Backfill |
| 2 | Transcript Mining |
| 3 | Episode Replay |
| 4 | Knowledge Consolidation |
| 5 | Index Refresh |
| 6 | Data Quality Reconciliation |
| 7 | Staleness Sweep |
| 8 | Refile Processing |
| 9 | Graph Maintenance |
| 10 | Consolidation Self-Check |
| 11 | Brain Sync |
When running on GitHub Actions, no PKB MCP server is available. Direct file tool operations (Bash, Glob, Grep, Read, Write, Edit) on markdown files are allowed. Do not edit files outside $AOPS_SESSIONS/ (except git commits/push).