メニュー
Load full memory context for the current task. Use automatically at the start of coding tasks to load corrections, decisions, and relevant background. Also use when switching topics mid-session.
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。
SOC 職業分類に基づく
Persist a memory with intelligence — classify, dedupe against existing memories, store with metadata, and link to related entries. Use when the user says "remember this", "save this", "don't forget", "keep this", "save it", or equivalent intent in any language. Do NOT fire for file saves, git commits, or code persistence — only knowledge/memory persistence.
Load full memory context for the current task. Use automatically at the start of coding tasks to load corrections, decisions, and relevant background. Also use when switching topics mid-session.
Run health diagnostics on the amem memory database. Use when the user wants to check memory health, embedding coverage, stale memories, or diagnose issues.
Export all memories as markdown or to a file. Use when the user wants to back up, share, or review their memories outside of amem.
List all memories, optionally filtered by type. Use when the user wants to browse their stored memories or see what's in a specific category.
Search memories quickly. Use when the user asks "what do you remember about", "recall", "search memory", or wants to find past decisions, corrections, or context.
| name | context |
| description | Load full memory context for the current task. Use automatically at the start of coding tasks to load corrections, decisions, and relevant background. Also use when switching topics mid-session. |
Load all relevant memory context for a topic or the current task.
Determine the topic from $ARGUMENTS or the current conversation context.
Execute this sequence:
a. Call memory_inject with the topic — surfaces corrections (MUST follow) and decisions (SHOULD follow)
b. Call reminder_check — show any overdue or upcoming reminders
c. Call memory_tier with action: "list", tier: "core" — load always-on context
d. If more context needed, call memory_context with the topic for broader background
Present the context naturally:
Apply corrections as absolute constraints — never violate them.
If Claude auto-memory is also active:
