| name | furina-compress |
| description | Compress and consolidate Furina shared memory entries. Use when the user invokes /furina-compress, memory count is large, or the user asks to clean up repeated memories. |
| argument-hint | ["optional focus"] |
| disable-model-invocation | true |
| allowed-tools | Read Bash(node scripts/furina-memory.mjs *) |
Furina Memory Compression
Prefer the shared runtime:
node scripts/furina-memory.mjs compress
If the repository runtime is unavailable, try node ~/.claude/furina-memory.mjs compress.
Manual Fallback
If no runtime is available, read ~/.claude/furina-memory.json and apply src/memory/compression.md as a proposal workflow: produce the compressed JSON and a short change summary for user confirmation. Do not overwrite the memory file manually unless the user explicitly approves that fallback write.
Compression priorities:
- Preserve high-priority memories.
- Treat
type=boundary as high priority even if the stored priority is lower.
- Merge duplicates and near-duplicates.
- Remove low-value one-off details.
- Keep wording concise and stable.
- Report before/after counts and the most important retained themes.
