| name | recall |
| description | Surface project memory, user profile, and knowledge into the current conversation.
Auto-fires in THREE scenarios:
1. EXPLICIT: User says /recall, "what do we know about [topic/project]", "remind me
where we left off", "how does [thing] work", "any gotchas with [thing]",
"what's the status of [project]", "catch me up on [project]". Also fires on
"what have we learned about [topic]" or when a user references a project/topic
and seems to expect prior knowledge.
2. CONVERSATION START: At the beginning of every new conversation, BEFORE the user
asks for anything, proactively load context. This means:
- Read the user profile node to remember who they are and how they like to work
- Check for active projects with stale threads or overdue P0s
- If the user's first message references a specific project or topic, load that
context immediately
- If the user's first message is a greeting or general start, load the dashboard
overview and surface anything that needs attention
- Apply user preferences from the profile to your behavior immediately
3. CONTEXTUAL: Mid-conversation when the user mentions a project, person, or topic
that has memory — surface the relevant knowledge without being asked. Don't
dump the full recall output, just weave in the relevant context naturally.
Example: user mentions "Acme" → you know their procurement requires 3 quotes
because it's in memory → mention it if relevant to the current discussion.
|
See commands/recall.md for the full workflow.
Auto-recall behavior (NEW in v4)
The biggest change: /recall no longer waits for the user to ask. It runs
automatically at conversation start and contextually during conversation.
Conversation start (auto-recall)
When a new conversation begins:
-
Always load the user profile first (user node — ~/Documents/Claude/memory/user.md)
- Apply known preferences immediately (communication style, response length, etc.)
- This happens silently — don't announce "I've loaded your preferences"
-
Assess the first message:
- If it references a specific project → full project recall (show it)
- If it's a greeting or general start → load dashboard, surface attention items
- If it's a direct question or task → load relevant context silently, weave in
-
Proactive health surfacing:
- P0 actions overdue by 3+ days → mention them
- Stale threads (3+ sessions without progress) → flag them
- Dormant nodes coming back to life → note the gap
- Recent cross-project signals → surface if relevant
-
Output for auto-recall on greeting/start:
Welcome back. Here's what needs attention:
**Overdue**
- [P0 action] in [node] (due [date])
**Stale**
- [thread] in [node] — no progress in [count] sessions
**Recent**
- Last worked on [node] ([date]): [1-line summary]
Which project are we picking up today?
Keep this SHORT. Don't dump the full dashboard — just the attention items.
If nothing needs attention, keep it even shorter: "All clear — what are we working on?"
Contextual recall (mid-conversation)
When the user mentions a project, person, or topic that has memory:
- Don't dump a recall block. Weave relevant knowledge into your response naturally.
- Example: User says "I need to send the proposal to Acme" → you know from memory
their procurement needs 3 quotes → say "Heads up — Acme's procurement requires
3 vendor quotes even for renewals, so factor in 2 weeks lead time."
- Only surface knowledge that's RELEVANT to what they're doing right now.
- This is the difference between a tool and a brain: a brain connects context
automatically, it doesn't wait to be queried.
Legacy behavior (still supported)
- Project reference → single-project recall with knowledge entries prominent
- Person name → cross-project person profile
- Topic/keyword → topic-based knowledge recall
- "Where did we leave off" → most recent nodes, last session's open threads
- No context available → say so, ask if they want to share
- Knowledge entries surfaced prominently — they're the most valuable recall content