---

name: og-compose description: Compose full memory context from oGMemory (profile, archives, session state, working set)

oGMemory Compose

Retrieve full composed memory context from oGMemory, including profile, archive history, session state, and relevant working set items.

When to use

• User asks to search or look up past conversations
• User asks "what did we discuss about X"
• User wants to recall previous decisions or context
• User types /og-compose

Instructions

1. Take the user's query from the arguments or infer it from the conversation context
2. Call the compose API and extract all context layers:

curl -s -X POST "${OG_MEMORY_URL:-http://localhost:8090}/api/v1/compose" \
  -H "Content-Type: application/json" \
  -d "{\"prompt\":\"<query>\",\"accountId\":\"${OG_MEMORY_ACCOUNT_ID:-acct-demo}\",\"userId\":\"${OG_MEMORY_USER_ID:-u-claude}\",\"agentId\":\"claude-code\"}" \
  | python3 -c "
import json, sys
data = json.load(sys.stdin)
parts = []
for key, label in [('identityContext','Profile'),('episodicContext','Archives'),('sessionContext','Session'),('retrievedEvidence','Working Set')]:
    val = (data.get(key) or '').strip()
    if val: parts.append(f'## {label}\n{val}')
if parts: print('\n\n'.join(parts))
else: print('No relevant context found.')
"

3. Present the results to the user in a clear format
4. If no results, suggest alternative queries

Examples

/og-compose marketplace plugin setup
/og-compose how did we fix the AGFS mount issue
/og-compose openclaw context engine