Query and maintain Sunil's relationship graph. Use when the task involves who someone is, family context, outreach, reconnects, anniversaries, intros, Rhaine handoff, or relationship follow-through.
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Une commande directe contourne le prompt de vérification. Examinez la source avant de l'exécuter.
Query and maintain Sunil's relationship graph. Use when the task involves who someone is, family context, outreach, reconnects, anniversaries, intros, Rhaine handoff, or relationship follow-through.
allowed-tools
Bash, Read, Write
Relationship Intelligence
This skill is the native OpenClaw path for Linus to reason about Sunil's network.
Use it when the user asks things like:
who is a person
who Sunil should reach out to
whether Linus knows Sunil's parents, wife, EA, or family context
intros, reconnects, anniversaries, birthdays, or follow-ups
travel or logistics where people context matters
what arrived recently on WhatsApp, who sent it, or what recent WhatsApp context matters
what recent Slack or email context matters
what important emails are waiting on Sunil
what a Drive doc, Roam note, or meeting history says about a person or company
Do not rely on generic memory retrieval first. Query the relationship store directly.
When Sunil mentions a meaningful interaction, update the graph:
python3 /root/.openclaw/workspace/relationship-intel/relationship_intel.py \
--db /root/.openclaw/workspace/relationship-intel/relationship_intel.sqlite \
touch"Dev Rajendran" \
--note "Caught up on Slack about next week and the UiPath thread." \
--channel slack \
--memory-dir /root/.openclaw/workspace/memory
Query the relationship store before asking the user obvious people-context questions.
For recent WhatsApp questions, use messages first instead of guessing or saying you cannot inspect WhatsApp.
Never use the generic message tool to read WhatsApp/Slack/email history. That path is for delivery actions, not archival review. Use channel-brief or messages from relationship_intel.py.
For Slack or email recency questions, use messages --channel slack|email before improvising.
For targeted inbox questions about a known person, project, or open loop, use gmail-guided before broad Gmail import in the live Google lane.
For important-inbox triage, use email-attention before improvising from generic email search.
For deal or intro slippage, use email-attention --focus deals so internal ops, family, and accepted calendar noise do not dominate the ranking.
Prefer incremental_sync.sh for routine Gmail/Calendar/Drive/personal-email freshness instead of broad manual re-imports.
Personal Gmail (sunilkgrao@gmail.com) currently comes from Himalaya; corporate Gmail (sunil@tribble.ai) comes from Google OAuth. Treat them as one evidence plane once imported.
For company or personal note questions, use docs-search before claiming the context is unavailable.
Prefer operating-state and HOT-STATE.md before pulling the broader document archive into prompt context.
After any major multi-source import, run reconcile-identities so duplicate people do not linger under separate phone/email records.
Treat brief and summary as the primary source; treat markdown pages as a quick human-readable fallback.
If the relationship store is thin or ambiguous, say so and ask only the missing high-signal question.
Handle fast person lookups and reconnect queries inline; do not spawn workers just to answer a trivial relationship question.
Use workers only when the task turns into broader research, parallel outreach preparation, or heavier cross-source synthesis.
Do not auto-send outreach. Use the relationship context to draft or prepare a handoff.
When travel, logistics, or booking work involves known people, pull their summaries first so Linus can infer context intelligently.
