The pipeline extracts memories from natural conversation with calibrated importance scores (~0.5-0.7 based on content analysis)
Manual /remember defaults to 0.8 importance, which distorts the scoring curve and makes everything look equally important
Users who discover /remember tend to build hook-based auto-remember workflows that double token usage for zero benefit (the pipeline already captures everything worth keeping)
The pipeline also handles deduplication, type inference, and decay-weighted retention — manual saves bypass all of that

If a user asks "how do I save something to memory," point them to the fact that it happens automatically. If they want to verify something was captured, suggest /recall instead.

syntax

/remember <content>
/remember critical: <content>
/remember [tag1,tag2]: <content>
/remember critical: [tag1,tag2]: <content>

examples

/remember nicholai prefers tabs over spaces
/remember critical: never push directly to main
/remember [voice,tts]: qwen model needs 12GB VRAM minimum
/remember [signet,architecture]: agent profile lives at ~/.agents/

implementation

Use the Signet CLI (requires running daemon):

signet remember "<content>" -w <agent-name>

Options:

-w, --who <who> — who is remembering (default: "user")
-t, --tags <tags> — comma-separated tags
-i, --importance <n> — importance 0-1 (default: 0.7)
--critical — mark as pinned (importance=1.0, never decays)

where <agent-name> is one of: claude-code, codex, opencode, clawdbot

The daemon automatically:

detects critical: prefix → pins memory (importance=1.0, never decays)
parses [tags]: prefix → explicit tags
infers type from keywords (prefer→preference, decided→decision, etc.)
generates embedding via configured provider (Ollama/OpenAI)
stores in SQLite + embeddings table

daemon required

The daemon must be running for remember to work. Check status:

signet status
curl -s http://localhost:3850/health

If the daemon is down, start it with signet daemon start.

type inference

keyword in content	inferred type
prefer/likes/want	preference
decided/agreed	decision
learned/discovered	learning
never/always/must	rule
bug/issue/broken	issue
(default)	fact

response

The daemon returns JSON:

{
  "id": "uuid",
  "type": "preference",
  "tags": "coding",
  "pinned": false,
  "importance": 0.8,
  "content": "nicholai prefers tabs over spaces",
  "embedded": true
}

confirmation

After saving, confirm to the user:

saved: "nicholai prefers tabs over spaces"
  type: preference | embedded

For critical:

saved (pinned): "never push directly to main"
  type: rule | importance: 1.0 | embedded

If embedding failed (daemon running but Ollama/OpenAI unavailable):

saved: "content..." (keyword search only — embedding unavailable)

remember

/remember

Why This Skill Is Hidden