memory

name	memory
description	Recall and store long-term memory in Codex using the memory-lancedb-pro MCP tools. Use when the user asks what happened before, wants to continue prior work, asks to remember a fix or preference, or types `@memory`.
metadata	{"short-description":"Recall and store long-term memory for this repo"}

Memory

Use this skill to make persistent memory feel like a first-class Codex workflow.

Explicit Invocation

• Typing $memory explicitly invokes this skill.
• In the Codex app, enabled skills also appear in the slash command list, so /memory should activate this skill in a new thread after reload.
• Treat plain text @memory or @记忆 as a user-facing alias for this skill even though it is not a native app-backed @ chip.

Default Workflow

1. Before broad repo exploration, call memory_recall if prior fixes, decisions, or preferences may matter.
2. Use scope project:new-project unless there is a better repo-specific slug in the current task.
3. After a durable fix, decision, or stable preference, prefer memory_extract_and_store with a short summary transcript.
4. Use memory_store only for a single short atomic memory.
5. Use memory_health if recall or writes fail.

Fast Mappings

• @memory recall <query>: call memory_recall
• @memory store <text>: call memory_store
• @memory extract <summary>: call memory_extract_and_store
• @memory list: call memory_list
• @memory stats: call memory_stats
• @memory health: call memory_health
• @memory forget <id>: call memory_forget

Writing Rules

• Prefer concise, reusable memories.
• Split facts and rules into separate memories when both matter.
• Do not store secrets, API keys, tokens, or private credentials.