obtain

Load memory entries as active session context.

Instructions

1. Parse $ARGUMENTS:
   • project:tagname → pass --tag tagname
   • plain word → pass --keyword WORD
   • empty → load all (may be large)
2. Run:

uv run skills/obtain/scripts/load_memory.py --keyword KEYWORD
# or
uv run skills/obtain/scripts/load_memory.py --tag TAGNAME

3. Read ALL of the output carefully and treat it as active context for this conversation.
4. Acknowledge to the user:
   • How many memory entries were loaded
   • The date range they cover
   • Tags represented
   • A 2–3 sentence synthesis of the most important points
5. If the script prints a CONTEXT_WARNING (>10 entries), relay it and ask the user to narrow the search.
6. If no entries match, say so and suggest /find-memories to browse available memories.

Active Context Rules

After loading, you MUST:

• Reference these memories when answering related questions
• Prefer decisions documented in memory over your defaults when there is a conflict
• Mention "Based on your saved context..." when using a memory to inform an answer

Storage

• Reads from ~/.claude/memory/ by default
• Respects CLAUDE_MEMORY_DIR when set