| name | recall |
| description | Recall project context or answer questions about history and decisions. Use before starting work, when investigating unfamiliar code, or asking "What was decided about X?" / "Why did we choose Y?" |
| allowed-tools | ["ToolSearch","mcp__watercooler__watercooler_smart_query"] |
Recall
Query: $ARGUMENTS
Framing Detection
After stripping flags (e.g. --raw), inspect the remaining text:
- Question framing if text ends with
? OR starts with a question word:
What, Why, How, Who, When, Where, Which, Is, Are, Was, Were, Did, Does, Can, Should
- Context framing otherwise (topic or task description)
Steps
-
Load MCP tool:
ToolSearch: select:mcp__watercooler__watercooler_smart_query
-
Execute query (scope to current repo with code_path):
mcp__watercooler__watercooler_smart_query(query="<$ARGUMENTS minus flags>", code_path="<repo root>")
-
Present results based on framing:
Question framing — answer the question:
- Direct answer to the question
- Evidence — cite specific entries by entry_id
- Confidence and tier used (T1/T2/T3)
- Escalation note if tier escalated beyond T1
- Suggested follow-ups if answer is incomplete or uncertain
Context framing — summarize relevant context:
- Prior decisions related to this topic
- Relevant patterns or implementations
- Known issues or gotchas
- Related threads for deeper reading
- Tier used and escalation reason
-
Handle empty results:
- Suggest alternative search terms
- Note whether memory backends are configured
-
Raw output (if --raw flag present):
- Append full JSON response after the summary
Example Invocations
/recall authentication flow — context before implementing auth
/recall What was decided about the config system? — direct question
/recall Why did we choose markdown for threads? — direct question
/recall --raw branch parity sync — raw JSON output
/recall Who implemented the branch parity feature? — attribution lookup