[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M77M823SX4QWNHM9AFQYYMQQ":3,"skill-same:01M77M823SX4QWNHM9AFQYYMQQ":74},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":53,"commitSha":54,"skillDirectory":55,"forksCount":14,"repositoryHtmlUrl":56,"repositoryHomepageUrl":6,"ownerHtmlUrl":57,"ownerBlogUrl":58,"ownerDescription":59,"updatedAt":60,"install":61,"resources":73},"01M77M823SX4QWNHM9AFQYYMQQ","cold-start",null,"Day-one data bootstrapping for a new brain. Sequences the highest-leverage\ndata sources to go from empty brain to useful brain in one session. Uses\nClawVisor for safe credential handling — the agent never holds raw API keys.\nCovers Gmail import, calendar sync, contacts seeding, X\u002FTwitter archive,\nconversation imports, and file archives.\nUse when a user has just finished gbrain setup and asks \"now what?\"","singhvedant\u002Ftbrain","singhvedant","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F71930695?v=4","01M77M5AXVZ7M5XWPANT3J9NS5","Vedant Singh","01M77M5AZ21DXDMR8ZPN3CPR07",0,"MIT","https:\u002F\u002Fgithub.com\u002Fsinghvedant\u002Ftbrain\u002Fblob\u002Fmain\u002Fskills\u002Fcold-start\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: cold-start\nversion: 1.0.0\ndescription: |\n  Day-one data bootstrapping for a new brain. Sequences the highest-leverage\n  data sources to go from empty brain to useful brain in one session. Uses\n  ClawVisor for safe credential handling — the agent never holds raw API keys.\n  Covers Gmail import, calendar sync, contacts seeding, X\u002FTwitter archive,\n  conversation imports, and file archives.\n  Use when a user has just finished gbrain setup and asks \"now what?\"\ntriggers:\n  - \"cold start\"\n  - \"fill my brain\"\n  - \"bootstrap brain\"\n  - \"import my data\"\n  - \"day one\"\n  - \"get started\"\n  - \"what should I import first\"\n  - \"populate brain\"\n  - \"now what?\"\ntools:\n  - search\n  - query\n  - get_page\n  - put_page\n  - add_link\n  - add_timeline_entry\n  - sync_brain\nmutating: true\nwrites_pages: true\nwrites_to:\n  - people\u002F\n  - companies\u002F\n  - meetings\u002F\n  - daily\u002F\n  - media\u002F\n  - conversations\u002F\n  - sources\u002F\n---\n\n# Cold Start — Day-One Brain Bootstrapping\n\nYou have a working brain. Search works. Now what?\n\nAn empty brain is a static database. A brain with your email history, calendar,\ncontacts, conversations, and social media is a **live context membrane** that makes\nevery future interaction smarter. This skill sequences the highest-leverage data\nsources to get you from zero to useful in one session.\n\n## Contract\n\n- Every import phase is gated on user consent (ask-user pattern) before proceeding.\n- **Google\u002Fsocial API access goes through ClawVisor.** The agent never holds raw OAuth\n  tokens or API keys. This is a safety requirement, not a preference. ClawVisor vaults\n  credentials, enforces task-scoped authorization, logs every API call, and requires\n  human approval for destructive operations. If the user doesn't want ClawVisor, the\n  only safe alternative is offline file exports (Google Takeout, Twitter archive download).\n- Each phase is independently valuable — the user can stop after any phase and still\n  have a useful brain.\n- Progress is tracked in `~\u002F.gbrain\u002Fcold-start-state.json` so interrupted sessions\n  can resume.\n- Entity detection and cross-linking run on every import, not as a separate pass.\n\n## Prerequisites\n\n- GBrain installed and initialized (`gbrain doctor --json` all green)\n- Brain repo cloned and synced\n- Agent has terminal access and can run `gbrain` CLI commands\n\n## The Priority Stack\n\nData sources ranked by **information density × ease of import**:\n\n| Priority | Source | Why | Time | Pages Created |\n|----------|--------|-----|------|---------------|\n| 1 | Existing markdown\u002FObsidian | Highest density — it's already structured | 5 min | 100s-1000s |\n| 2 | Google Contacts | Seeds the people\u002F directory — names, emails, companies | 10 min | 50-500 |\n| 3 | Google Calendar (90 days) | Meeting history with attendee context | 15 min | 30-90 |\n| 4 | Gmail (recent threads) | Relationship context, active threads, org chart signals | 20 min | 50-200 |\n| 5 | Conversations (ChatGPT\u002FClaude exports) | Your thinking, questions, mental models | 15 min | 10-100 |\n| 6 | X\u002FTwitter archive | Your public positions, takes, engagement patterns | 20 min | 30-365 |\n| 7 | File archives (Dropbox\u002FDrive\u002Flocal) | Historical documents, old writing, photos | 30+ min | varies |\n| 8 | Meeting transcripts (Circleback\u002Fetc.) | Deep relationship context from recorded calls | 20 min | 10-50 |\n\n## Phase 0: ClawVisor Setup (Required for API Access)\n\n> **Safety boundary:** An AI agent with raw OAuth tokens to your Gmail, Calendar,\n> and Contacts is an uncontrolled attack surface. One prompt injection, one\n> malicious tool call, and your entire Google account is exposed. ClawVisor\n> eliminates this risk class entirely.\n\n[ClawVisor](https:\u002F\u002Fclawvisor.com) is a credential gateway that sits between the\nagent and your APIs. The agent never sees your credentials — ClawVisor injects\nthem at request time, enforces policies, and logs everything.\n\n**What ClawVisor gives you:**\n- **Credential vaulting** — agent sees shadow tokens, never real secrets\n- **Task-scoped authorization** — each workflow declares exactly what it needs\n- **Audit trail** — every API call logged with metadata (who, what, when)\n- **Human approval gates** — destructive operations (send email, modify calendar)\n  require your explicit approval\n- **Multi-service** — Gmail, Calendar, Contacts, Drive, GitHub, iMessage from one gateway\n- **Revocation** — disable the agent's access in one click, no token rotation needed\n\n**Setup (15 min):**\n1. Sign up at [app.clawvisor.com](https:\u002F\u002Fapp.clawvisor.com)\n2. Create an agent in the dashboard, copy the agent token\n3. Set environment variables:\n   ```bash\n   gbrain config set clawvisor_url \"https:\u002F\u002Fapp.clawvisor.com\"\n   gbrain config set clawvisor_agent_token \"\u003Ctoken>\"\n   ```\n4. Activate Google services (Gmail, Calendar, Contacts) in the dashboard\n5. Create a standing task with expansive scope:\n   > \"Full brain bootstrapping: read emails, calendar events, and contacts to\n   > populate knowledge base. List, read, and search across all connected accounts.\"\n6. Save the standing task ID:\n   ```bash\n   gbrain config set clawvisor_task_id \"\u003Ctask_id>\"\n   ```\n\n**Critical scoping rule:** Be expansive in task purposes. \"Email triage\" gets\nrejected by intent verification. \"Full executive assistant email management\nincluding inbox triage, searching by any criteria, reading emails, tracking\nthreads\" works. The intent model uses the purpose to judge each request.\n\n### If the user declines ClawVisor\n\nDo NOT fall back to direct OAuth. Instead, skip Phases 2-4 (Contacts, Calendar,\nGmail) and proceed with offline-only imports:\n\n- **Phase 1** (markdown\u002FObsidian) — works without any API access\n- **Phase 5** (conversation exports) — works from downloaded JSON files\n- **Phase 6** (X\u002FTwitter) — works from downloaded archive\n- **Phase 7** (file archives) — works from local files\n- **Phase 8** (meeting transcripts) — works from exported transcripts\n\nTell the user:\n> \"No problem. We'll skip the Google imports for now and work with file-based\n> sources. You can set up ClawVisor anytime to unlock Contacts, Calendar, and\n> Gmail imports safely.\"\n\n**Do NOT offer direct OAuth as an alternative.** An agent holding raw Google\ntokens is a security liability. The skill should not teach agents to store\ncredentials they shouldn't have.\n\n## Phase 1: Existing Markdown \u002F Obsidian Import\n\n**The highest-leverage first import.** If the user already has a notes system, this\nis hundreds or thousands of structured pages ready to go.\n\n### Discovery\n\n```bash\necho \"=== Markdown Repository Discovery ===\"\nfor dir in \u002Fdata\u002F* ~\u002Fgit\u002F* ~\u002FDocuments\u002F* ~\u002Fnotes\u002F* ~\u002Fobsidian\u002F* 2>\u002Fdev\u002Fnull; do\n  if [ -d \"$dir\" ]; then\n    md_count=$(find \"$dir\" -name \"*.md\" -not -path \"*\u002Fnode_modules\u002F*\" \\\n      -not -path \"*\u002F.git\u002F*\" -not -path \"*\u002F.obsidian\u002F*\" 2>\u002Fdev\u002Fnull | wc -l | tr -d ' ')\n    if [ \"$md_count\" -gt 5 ]; then\n      total_size=$(du -sh \"$dir\" 2>\u002Fdev\u002Fnull | cut -f1)\n      echo \"  $dir ($total_size, $md_count .md files)\"\n    fi\n  fi\ndone\n```\n\n### Import\n\n```bash\n# For Obsidian vaults, use the migrate skill for proper wikilink handling\ngbrain migrate --from obsidian --path \u002Fpath\u002Fto\u002Fvault\n\n# For plain markdown directories\ngbrain import \u002Fpath\u002Fto\u002Fdir --no-embed --workers 4\n\n# Verify\ngbrain stats\ngbrain search \"\u003Ctopic from the imported data>\"\n```\n\n### Post-import\n\n- Run link extraction: `gbrain extract links --source db`\n- Run timeline extraction: `gbrain extract timeline --source db`\n- Start embeddings: `gbrain embed --stale` (runs in background)\n\n> **Track progress:**\n> ```bash\n> echo '{\"phase_1_complete\": true, \"pages_imported\": N}' > ~\u002F.gbrain\u002Fcold-start-state.json\n> ```\n\n## Phase 2: Google Contacts → People Pages\n\n**Seeds the people\u002F directory.** Every person in your contacts becomes a brain page\nwith name, email, phone, company, and notes. This is the foundation that all other\nimports build on — when Gmail references \"john@acme.com\", the brain already knows\nwho John is.\n\n### Via ClawVisor\n\n```javascript\n\u002F\u002F Fetch all contacts\nconst contacts = await clawvisor('google.contacts', 'list_contacts', {\n  limit: 1000,\n  fields: 'names,emailAddresses,phoneNumbers,organizations,biographies'\n});\n```\n\n### Via direct Google People API\n\n```bash\ncurl -s -H \"Authorization: Bearer $GOOGLE_TOKEN\" \\\n  \"https:\u002F\u002Fpeople.googleapis.com\u002Fv1\u002Fpeople\u002Fme\u002Fconnections?personFields=names,emailAddresses,phoneNumbers,organizations,biographies&pageSize=1000\"\n```\n\n### Processing rules\n\nFor each contact:\n1. **Filter out noise** — skip contacts with no name, no email, or that are clearly\n   automated (noreply@, no-reply@, support@, notifications@)\n2. **Check brain first** — `gbrain search \"name\"` to avoid duplicates\n3. **Create people\u002F page** with:\n   - Name, email(s), phone(s), company, title\n   - Source attribution: `[Source: Google Contacts, YYYY-MM-DD]`\n   - Any notes from the contact as initial context\n4. **Link to company** — if the contact has an organization, create\u002Fupdate the\n   company page and link the person to it\n\n### Quality gate\n\nAfter importing 5 contacts, pause and show the user a sample page. Ask:\n> \"Here's what a contact page looks like. Want me to continue with the rest, or\n> adjust the format first?\"\n\n## Phase 3: Google Calendar (Last 90 Days)\n\n**Meeting history with attendee context.** Calendar events reveal who the user meets\nwith, how often, and in what context. Combined with contacts, this builds a rich\nrelationship map.\n\n### Fetch events\n\n```javascript\n\u002F\u002F Via ClawVisor — query ALL calendar accounts\nconst accounts = ['primary@gmail.com', 'work@company.com'];\nfor (const account of accounts) {\n  const events = await clawvisor(`google.calendar:${account}`, 'list_events', {\n    timeMin: new Date(Date.now() - 90 * 86400000).toISOString(),\n    timeMax: new Date().toISOString(),\n    singleEvents: true,\n    orderBy: 'startTime'\n  });\n}\n```\n\n### Brain structure\n\nFollow the three-tier calendar architecture:\n```\nbrain\u002Fdaily\u002Fcalendar\u002F\n├── calendar-log.md              ← compiled truth (patterns, key people)\n├── YYYY\u002F\n│   ├── YYYY-MM.md               ← monthly summary\n│   └── YYYY-MM-DD.md            ← daily event log\n```\n\n### Entity enrichment\n\nFor each event with attendees:\n1. Look up each attendee in the brain (they should exist from Phase 2)\n2. Add a timeline entry to their page: met at [event title] on [date]\n3. If an attendee has no brain page and appears in 3+ events, create one\n4. Link attendees who appear in the same meeting\n\n## Phase 4: Gmail (Recent Threads)\n\n**Relationship context and active threads.** Email reveals organizational\nrelationships, ongoing conversations, and communication patterns.\n\n### Strategy: Smart sampling, not bulk import\n\nDon't import every email. Import the **signal**:\n\n1. **Sent mail (last 30 days)** — who the user actively communicates with\n2. **Starred\u002Fimportant emails** — user-curated signal\n3. **Threads with 3+ replies** — active conversations worth tracking\n4. **Emails from people already in the brain** — enrichment, not cold import\n\n### Processing\n\nFor each email thread:\n1. **Entity detection** — extract people, companies mentioned\n2. **Update people pages** — add communication context to timeline\n3. **Create meeting pages** — if the email is a meeting summary or follow-up\n4. **Skip noise** — newsletters, automated notifications, marketing\n\n### Filtering rules\n\n**Auto-skip (never import):**\n- noreply@, no-reply@, notifications@, support@, mailer-daemon@\n- Unsubscribe-heavy senders (marketing)\n- GitHub\u002FJira\u002FLinear notification emails\n- Calendar invites (already captured in Phase 3)\n\n**Always import:**\n- Direct emails from people in the brain\n- Starred\u002Fflagged emails\n- Emails the user sent (their words are highest-value signal)\n\n## Phase 5: Conversation Exports (ChatGPT \u002F Claude \u002F Perplexity)\n\n**Your thinking, captured.** AI conversation exports reveal what the user\nwas researching, building, and thinking about. This is original thinking\npreserved in dialog form.\n\n### Supported formats\n\n- **ChatGPT:** Settings → Data Controls → Export → `conversations.json`\n- **Claude:** Download from claude.ai conversation history\n- **Perplexity:** Export from settings\n\n### Processing\n\nFor each conversation:\n1. **Assess significance** (1-5 scale):\n   - 1 = Pure utility (how-tos, quick lookups) → skip or minimal page\n   - 2 = Minor context → 1-paragraph note\n   - 3 = Notable (reveals interests, building something) → full page\n   - 4 = Important (deep personal processing, strategic thinking) → rich page\n   - 5 = Defining (identity work, breakthrough insights) → full treatment\n2. **Extract entities** — people, companies, concepts discussed\n3. **Capture original thinking** — the user's exact phrasing is the signal.\n   Never paraphrase.\n4. **File by primary subject** — not in a \"conversations\u002F\" dump. A conversation\n   about a person goes to people\u002F, about a concept goes to concepts\u002F, etc.\n\n### Quality rule\n\nOnly import conversations rated 3+. The brain is for signal, not noise.\n\n## Phase 6: X\u002FTwitter Archive\n\n**Your public positions and engagement patterns.** Twitter reveals what the user\nthinks, who they engage with, and what ideas they're developing publicly.\n\n### Data sources\n\n1. **Twitter data export** (Settings → Your Account → Download Archive)\n   - Contains all tweets, likes, DMs, bookmarks\n2. **Live API** (if available) — recent tweets and engagement\n3. **Bookmarks** — curated signal, high value\n\n### Brain structure\n\n```\nbrain\u002Fmedia\u002Fx\u002F{handle}\u002F\n├── x-log.md                     ← compiled truth (themes, voice, key threads)\n├── daily\u002FYYYY-MM-DD.md          ← daily tweet log\n├── monthly\u002FYYYY-MM.md           ← monthly rollup\n└── bookmarks\u002F                   ← saved\u002Fbookmarked content\n```\n\n### Processing\n\n- **Original tweets** → capture with full context, extract entities\n- **Quote tweets** → capture the user's commentary + the source tweet\n- **Threads** → reconstruct as a single narrative\n- **Bookmarks** → high-signal curation, import with tags\n- **Likes** — low signal, skip unless the user wants them\n\n## Phase 7: File Archives\n\n**Historical documents, old writing, photos with metadata.** This is the long tail —\nless structured but potentially very high value (old journals, letters, early writing).\n\nDelegate to the `archive-crawler` skill. It handles:\n- Crawling directory structures\n- Filtering for high-value content (user's own writing, not installers)\n- Text extraction from PDFs, images (OCR), documents\n- Entity extraction and brain page creation\n\n> **Safety gate:** Archive crawling can be slow and create many pages. Always start\n> with a scan-only pass:\n> ```bash\n> gbrain archive-crawler --scan-only --path \u002Fpath\u002Fto\u002Farchive\n> ```\n> Show the user the manifest before proceeding with full ingestion.\n\n**Supported sources:**\n- Local directories (Dropbox sync folder, Google Drive, old hard drives)\n- Cloud storage (Backblaze B2, S3) via mounted paths\n- Email archives (PST, mbox, EML, Google Takeout)\n- Data exports (LinkedIn, Facebook, etc.)\n\n## Phase 8: Meeting Transcripts\n\n**Deep relationship context from recorded calls.** If the user has a meeting\nrecording service (Circleback, Otter, Fireflies, Read.ai), import recent\ntranscripts.\n\nDelegate to `meeting-ingestion` skill. Key rules:\n- Always pull the **complete transcript**, not just the AI summary\n- Entity propagation is MANDATORY — every attendee gets a timeline update\n- A meeting is NOT fully ingested until all entity pages are updated\n\n## Post-Bootstrap Checklist\n\nAfter completing available phases:\n\n1. **Verify brain health:**\n   ```bash\n   gbrain doctor --json\n   gbrain stats\n   ```\n\n2. **Test retrieval:**\n   ```bash\n   gbrain query \"who do I meet with most often?\"\n   gbrain query \"what am I working on?\"\n   gbrain search \"\u003Cperson from contacts>\"\n   ```\n\n3. **Set up live sync** (if not already):\n   - Calendar: daily cron\n   - Email: periodic sweep (4-8 hours)\n   - X: daily ingest\n   - Brain repo: `gbrain sync --repo \u003Cpath>` every 5-30 minutes\n\n4. **Track state:**\n   ```json\n   \u002F\u002F ~\u002F.gbrain\u002Fcold-start-state.json\n   {\n     \"started\": \"2026-01-15T10:00:00Z\",\n     \"credential_gateway\": \"clawvisor\",\n     \"phases_completed\": [1, 2, 3, 4],\n     \"phases_skipped\": [6, 7],\n     \"total_pages_created\": 847,\n     \"total_entities_linked\": 1203,\n     \"next_phase\": 5\n   }\n   ```\n\n5. **Tell the user what to do next:**\n   > \"Your brain has N pages across people, calendar, email, and conversations.\n   > Live sync is configured for [sources]. From here:\n   > - The **signal-detector** captures entities from every conversation\n   > - The **briefing** skill can compile daily context\n   > - The **executive-assistant** pattern handles email triage\n   > - Say 'enrich [person]' to deep-dive any contact\"\n\n## Anti-Patterns\n\n- **Giving the agent raw OAuth tokens.** This is the #1 anti-pattern. An agent with\n  raw Gmail\u002FCalendar tokens is an uncontrolled attack surface — one prompt injection\n  and your entire Google account is exposed. Use ClawVisor. If the user declines\n  ClawVisor, skip to offline imports. Never offer direct OAuth as a fallback.\n- **Bulk importing everything without filtering.** The brain is for signal, not noise.\n  Filter out automated senders, marketing emails, utility conversations.\n- **Importing without entity cross-linking.** Every import should detect entities and\n  update existing brain pages. Isolated imports don't compound.\n- **Not gating on user consent.** Every phase should be presented as a choice. The user\n  may not want their DMs or therapy conversations imported.\n- **Importing everything at significance 1.** Not every conversation is worth a brain\n  page. Use the significance scale and skip utility content.\n- **Creating people pages for automated senders.** Sentry, GitHub notifications,\n  newsletter platforms are not people. Filter by the rules in Phase 4.\n\n## Resume Protocol\n\nIf the session is interrupted:\n\n1. Read `~\u002F.gbrain\u002Fcold-start-state.json`\n2. Skip completed phases\n3. Resume from `next_phase`\n4. The user doesn't have to repeat credential setup or re-import completed sources\n\n## Output Format\n\nAfter each phase:\n\n```\nPHASE N COMPLETE: [source name]\n================================\n\nPages created: N\nPages updated: N\nEntities linked: N\nTime elapsed: N min\n\nSample pages:\n- people\u002Fjane-smith.md (created — 3 emails, 5 meetings)\n- companies\u002Facme-corp.md (updated — 2 new employees linked)\n\nNext: Phase N+1 — [description]. Ready to proceed?\n```\n\n## Tools Used\n\n- `search` — check for existing pages before creating\n- `query` — hybrid search for entity deduplication\n- `get_page` — read existing pages for merge decisions\n- `put_page` — create and update brain pages\n- `add_link` — cross-reference entities\n- `add_timeline_entry` — record events on entity timelines\n- `sync_brain` — sync changes to the index after each phase\n","# Cold Start — Day-One Brain Bootstrapping\n\nYou have a working brain. Search works. Now what?\n\nAn empty brain is a static database. A brain with your email history, calendar,\ncontacts, conversations, and social media is a **live context membrane** that makes\nevery future interaction smarter. This skill sequences the highest-leverage data\nsources to get you from zero to useful in one session.\n\n## Contract\n\n- Every import phase is gated on user consent (ask-user pattern) before proceeding.\n- **Google\u002Fsocial API access goes through ClawVisor.** The agent never holds raw OAuth\n  tokens or API keys. This is a safety requirement, not a preference. ClawVisor vaults\n  credentials, enforces task-scoped authorization, logs every API call, and requires\n  human approval for destructive operations. If the user doesn't want ClawVisor, the\n  only safe alternative is offline file exports (Google Takeout, Twitter archive download).\n- Each phase is independently valuable — the user can stop after any phase and still\n  have a useful brain.\n- Progress is tracked in `~\u002F.gbrain\u002Fcold-start-state.json` so interrupted sessions\n  can resume.\n- Entity detection and cross-linking run on every import, not as a separate pass.\n\n## Prerequisites\n\n- GBrain installed and initialized (`gbrain doctor --json` all green)\n- Brain repo cloned and synced\n- Agent has terminal access and can run `gbrain` CLI commands\n\n## The Priority Stack\n\nData sources ranked by **information density × ease of import**:\n\n| Priority | Source | Why | Time | Pages Created |\n|----------|--------|-----|------|---------------|\n| 1 | Existing markdown\u002FObsidian | Highest density — it's already structured | 5 min | 100s-1000s |\n| 2 | Google Contacts | Seeds the people\u002F directory — names, emails, companies | 10 min | 50-500 |\n| 3 | Google Calendar (90 days) | Meeting history with attendee context | 15 min | 30-90 |\n| 4 | Gmail (recent threads) | Relationship context, active threads, org chart signals | 20 min | 50-200 |\n| 5 | Conversations (ChatGPT\u002FClaude exports) | Your thinking, questions, mental models | 15 min | 10-100 |\n| 6 | X\u002FTwitter archive | Your public positions, takes, engagement patterns | 20 min | 30-365 |\n| 7 | File archives (Dropbox\u002FDrive\u002Flocal) | Historical documents, old writing, photos | 30+ min | varies |\n| 8 | Meeting transcripts (Circleback\u002Fetc.) | Deep relationship context from recorded calls | 20 min | 10-50 |\n\n## Phase 0: ClawVisor Setup (Required for API Access)\n\n> **Safety boundary:** An AI agent with raw OAuth tokens to your Gmail, Calendar,\n> and Contacts is an uncontrolled attack surface. One prompt injection, one\n> malicious tool call, and your entire Google account is exposed. ClawVisor\n> eliminates this risk class entirely.\n\n[ClawVisor](https:\u002F\u002Fclawvisor.com) is a credential gateway that sits between the\nagent and your APIs. The agent never sees your credentials — ClawVisor injects\nthem at request time, enforces policies, and logs everything.\n\n**What ClawVisor gives you:**\n- **Credential vaulting** — agent sees shadow tokens, never real secrets\n- **Task-scoped authorization** — each workflow declares exactly what it needs\n- **Audit trail** — every API call logged with metadata (who, what, when)\n- **Human approval gates** — destructive operations (send email, modify calendar)\n  require your explicit approval\n- **Multi-service** — Gmail, Calendar, Contacts, Drive, GitHub, iMessage from one gateway\n- **Revocation** — disable the agent's access in one click, no token rotation needed\n\n**Setup (15 min):**\n1. Sign up at [app.clawvisor.com](https:\u002F\u002Fapp.clawvisor.com)\n2. Create an agent in the dashboard, copy the agent token\n3. Set environment variables:\n   ```bash\n   gbrain config set clawvisor_url \"https:\u002F\u002Fapp.clawvisor.com\"\n   gbrain config set clawvisor_agent_token \"\u003Ctoken>\"\n   ```\n4. Activate Google services (Gmail, Calendar, Contacts) in the dashboard\n5. Create a standing task with expansive scope:\n   > \"Full brain bootstrapping: read emails, calendar events, and contacts to\n   > populate knowledge base. List, read, and search across all connected accounts.\"\n6. Save the standing task ID:\n   ```bash\n   gbrain config set clawvisor_task_id \"\u003Ctask_id>\"\n   ```\n\n**Critical scoping rule:** Be expansive in task purposes. \"Email triage\" gets\nrejected by intent verification. \"Full executive assistant email management\nincluding inbox triage, searching by any criteria, reading emails, tracking\nthreads\" works. The intent model uses the purpose to judge each request.\n\n### If the user declines ClawVisor\n\nDo NOT fall back to direct OAuth. Instead, skip Phases 2-4 (Contacts, Calendar,\nGmail) and proceed with offline-only imports:\n\n- **Phase 1** (markdown\u002FObsidian) — works without any API access\n- **Phase 5** (conversation exports) — works from downloaded JSON files\n- **Phase 6** (X\u002FTwitter) — works from downloaded archive\n- **Phase 7** (file archives) — works from local files\n- **Phase 8** (meeting transcripts) — works from exported transcripts\n\nTell the user:\n> \"No problem. We'll skip the Google imports for now and work with file-based\n> sources. You can set up ClawVisor anytime to unlock Contacts, Calendar, and\n> Gmail imports safely.\"\n\n**Do NOT offer direct OAuth as an alternative.** An agent holding raw Google\ntokens is a security liability. The skill should not teach agents to store\ncredentials they shouldn't have.\n\n## Phase 1: Existing Markdown \u002F Obsidian Import\n\n**The highest-leverage first import.** If the user already has a notes system, this\nis hundreds or thousands of structured pages ready to go.\n\n### Discovery\n\n```bash\necho \"=== Markdown Repository Discovery ===\"\nfor dir in \u002Fdata\u002F* ~\u002Fgit\u002F* ~\u002FDocuments\u002F* ~\u002Fnotes\u002F* ~\u002Fobsidian\u002F* 2>\u002Fdev\u002Fnull; do\n  if [ -d \"$dir\" ]; then\n    md_count=$(find \"$dir\" -name \"*.md\" -not -path \"*\u002Fnode_modules\u002F*\" \\\n      -not -path \"*\u002F.git\u002F*\" -not -path \"*\u002F.obsidian\u002F*\" 2>\u002Fdev\u002Fnull | wc -l | tr -d ' ')\n    if [ \"$md_count\" -gt 5 ]; then\n      total_size=$(du -sh \"$dir\" 2>\u002Fdev\u002Fnull | cut -f1)\n      echo \"  $dir ($total_size, $md_count .md files)\"\n    fi\n  fi\ndone\n```\n\n### Import\n\n```bash\n# For Obsidian vaults, use the migrate skill for proper wikilink handling\ngbrain migrate --from obsidian --path \u002Fpath\u002Fto\u002Fvault\n\n# For plain markdown directories\ngbrain import \u002Fpath\u002Fto\u002Fdir --no-embed --workers 4\n\n# Verify\ngbrain stats\ngbrain search \"\u003Ctopic from the imported data>\"\n```\n\n### Post-import\n\n- Run link extraction: `gbrain extract links --source db`\n- Run timeline extraction: `gbrain extract timeline --source db`\n- Start embeddings: `gbrain embed --stale` (runs in background)\n\n> **Track progress:**\n> ```bash\n> echo '{\"phase_1_complete\": true, \"pages_imported\": N}' > ~\u002F.gbrain\u002Fcold-start-state.json\n> ```\n\n## Phase 2: Google Contacts → People Pages\n\n**Seeds the people\u002F directory.** Every person in your contacts becomes a brain page\nwith name, email, phone, company, and notes. This is the foundation that all other\nimports build on — when Gmail references \"john@acme.com\", the brain already knows\nwho John is.\n\n### Via ClawVisor\n\n```javascript\n\u002F\u002F Fetch all contacts\nconst contacts = await clawvisor('google.contacts', 'list_contacts', {\n  limit: 1000,\n  fields: 'names,emailAddresses,phoneNumbers,organizations,biographies'\n});\n```\n\n### Via direct Google People API\n\n```bash\ncurl -s -H \"Authorization: Bearer $GOOGLE_TOKEN\" \\\n  \"https:\u002F\u002Fpeople.googleapis.com\u002Fv1\u002Fpeople\u002Fme\u002Fconnections?personFields=names,emailAddresses,phoneNumbers,organizations,biographies&pageSize=1000\"\n```\n\n### Processing rules\n\nFor each contact:\n1. **Filter out noise** — skip contacts with no name, no email, or that are clearly\n   automated (noreply@, no-reply@, support@, notifications@)\n2. **Check brain first** — `gbrain search \"name\"` to avoid duplicates\n3. **Create people\u002F page** with:\n   - Name, email(s), phone(s), company, title\n   - Source attribution: `[Source: Google Contacts, YYYY-MM-DD]`\n   - Any notes from the contact as initial context\n4. **Link to company** — if the contact has an organization, create\u002Fupdate the\n   company page and link the person to it\n\n### Quality gate\n\nAfter importing 5 contacts, pause and show the user a sample page. Ask:\n> \"Here's what a contact page looks like. Want me to continue with the rest, or\n> adjust the format first?\"\n\n## Phase 3: Google Calendar (Last 90 Days)\n\n**Meeting history with attendee context.** Calendar events reveal who the user meets\nwith, how often, and in what context. Combined with contacts, this builds a rich\nrelationship map.\n\n### Fetch events\n\n```javascript\n\u002F\u002F Via ClawVisor — query ALL calendar accounts\nconst accounts = ['primary@gmail.com', 'work@company.com'];\nfor (const account of accounts) {\n  const events = await clawvisor(`google.calendar:${account}`, 'list_events', {\n    timeMin: new Date(Date.now() - 90 * 86400000).toISOString(),\n    timeMax: new Date().toISOString(),\n    singleEvents: true,\n    orderBy: 'startTime'\n  });\n}\n```\n\n### Brain structure\n\nFollow the three-tier calendar architecture:\n```\nbrain\u002Fdaily\u002Fcalendar\u002F\n├── calendar-log.md              ← compiled truth (patterns, key people)\n├── YYYY\u002F\n│   ├── YYYY-MM.md               ← monthly summary\n│   └── YYYY-MM-DD.md            ← daily event log\n```\n\n### Entity enrichment\n\nFor each event with attendees:\n1. Look up each attendee in the brain (they should exist from Phase 2)\n2. Add a timeline entry to their page: met at [event title] on [date]\n3. If an attendee has no brain page and appears in 3+ events, create one\n4. Link attendees who appear in the same meeting\n\n## Phase 4: Gmail (Recent Threads)\n\n**Relationship context and active threads.** Email reveals organizational\nrelationships, ongoing conversations, and communication patterns.\n\n### Strategy: Smart sampling, not bulk import\n\nDon't import every email. Import the **signal**:\n\n1. **Sent mail (last 30 days)** — who the user actively communicates with\n2. **Starred\u002Fimportant emails** — user-curated signal\n3. **Threads with 3+ replies** — active conversations worth tracking\n4. **Emails from people already in the brain** — enrichment, not cold import\n\n### Processing\n\nFor each email thread:\n1. **Entity detection** — extract people, companies mentioned\n2. **Update people pages** — add communication context to timeline\n3. **Create meeting pages** — if the email is a meeting summary or follow-up\n4. **Skip noise** — newsletters, automated notifications, marketing\n\n### Filtering rules\n\n**Auto-skip (never import):**\n- noreply@, no-reply@, notifications@, support@, mailer-daemon@\n- Unsubscribe-heavy senders (marketing)\n- GitHub\u002FJira\u002FLinear notification emails\n- Calendar invites (already captured in Phase 3)\n\n**Always import:**\n- Direct emails from people in the brain\n- Starred\u002Fflagged emails\n- Emails the user sent (their words are highest-value signal)\n\n## Phase 5: Conversation Exports (ChatGPT \u002F Claude \u002F Perplexity)\n\n**Your thinking, captured.** AI conversation exports reveal what the user\nwas researching, building, and thinking about. This is original thinking\npreserved in dialog form.\n\n### Supported formats\n\n- **ChatGPT:** Settings → Data Controls → Export → `conversations.json`\n- **Claude:** Download from claude.ai conversation history\n- **Perplexity:** Export from settings\n\n### Processing\n\nFor each conversation:\n1. **Assess significance** (1-5 scale):\n   - 1 = Pure utility (how-tos, quick lookups) → skip or minimal page\n   - 2 = Minor context → 1-paragraph note\n   - 3 = Notable (reveals interests, building something) → full page\n   - 4 = Important (deep personal processing, strategic thinking) → rich page\n   - 5 = Defining (identity work, breakthrough insights) → full treatment\n2. **Extract entities** — people, companies, concepts discussed\n3. **Capture original thinking** — the user's exact phrasing is the signal.\n   Never paraphrase.\n4. **File by primary subject** — not in a \"conversations\u002F\" dump. A conversation\n   about a person goes to people\u002F, about a concept goes to concepts\u002F, etc.\n\n### Quality rule\n\nOnly import conversations rated 3+. The brain is for signal, not noise.\n\n## Phase 6: X\u002FTwitter Archive\n\n**Your public positions and engagement patterns.** Twitter reveals what the user\nthinks, who they engage with, and what ideas they're developing publicly.\n\n### Data sources\n\n1. **Twitter data export** (Settings → Your Account → Download Archive)\n   - Contains all tweets, likes, DMs, bookmarks\n2. **Live API** (if available) — recent tweets and engagement\n3. **Bookmarks** — curated signal, high value\n\n### Brain structure\n\n```\nbrain\u002Fmedia\u002Fx\u002F{handle}\u002F\n├── x-log.md                     ← compiled truth (themes, voice, key threads)\n├── daily\u002FYYYY-MM-DD.md          ← daily tweet log\n├── monthly\u002FYYYY-MM.md           ← monthly rollup\n└── bookmarks\u002F                   ← saved\u002Fbookmarked content\n```\n\n### Processing\n\n- **Original tweets** → capture with full context, extract entities\n- **Quote tweets** → capture the user's commentary + the source tweet\n- **Threads** → reconstruct as a single narrative\n- **Bookmarks** → high-signal curation, import with tags\n- **Likes** — low signal, skip unless the user wants them\n\n## Phase 7: File Archives\n\n**Historical documents, old writing, photos with metadata.** This is the long tail —\nless structured but potentially very high value (old journals, letters, early writing).\n\nDelegate to the `archive-crawler` skill. It handles:\n- Crawling directory structures\n- Filtering for high-value content (user's own writing, not installers)\n- Text extraction from PDFs, images (OCR), documents\n- Entity extraction and brain page creation\n\n> **Safety gate:** Archive crawling can be slow and create many pages. Always start\n> with a scan-only pass:\n> ```bash\n> gbrain archive-crawler --scan-only --path \u002Fpath\u002Fto\u002Farchive\n> ```\n> Show the user the manifest before proceeding with full ingestion.\n\n**Supported sources:**\n- Local directories (Dropbox sync folder, Google Drive, old hard drives)\n- Cloud storage (Backblaze B2, S3) via mounted paths\n- Email archives (PST, mbox, EML, Google Takeout)\n- Data exports (LinkedIn, Facebook, etc.)\n\n## Phase 8: Meeting Transcripts\n\n**Deep relationship context from recorded calls.** If the user has a meeting\nrecording service (Circleback, Otter, Fireflies, Read.ai), import recent\ntranscripts.\n\nDelegate to `meeting-ingestion` skill. Key rules:\n- Always pull the **complete transcript**, not just the AI summary\n- Entity propagation is MANDATORY — every attendee gets a timeline update\n- A meeting is NOT fully ingested until all entity pages are updated\n\n## Post-Bootstrap Checklist\n\nAfter completing available phases:\n\n1. **Verify brain health:**\n   ```bash\n   gbrain doctor --json\n   gbrain stats\n   ```\n\n2. **Test retrieval:**\n   ```bash\n   gbrain query \"who do I meet with most often?\"\n   gbrain query \"what am I working on?\"\n   gbrain search \"\u003Cperson from contacts>\"\n   ```\n\n3. **Set up live sync** (if not already):\n   - Calendar: daily cron\n   - Email: periodic sweep (4-8 hours)\n   - X: daily ingest\n   - Brain repo: `gbrain sync --repo \u003Cpath>` every 5-30 minutes\n\n4. **Track state:**\n   ```json\n   \u002F\u002F ~\u002F.gbrain\u002Fcold-start-state.json\n   {\n     \"started\": \"2026-01-15T10:00:00Z\",\n     \"credential_gateway\": \"clawvisor\",\n     \"phases_completed\": [1, 2, 3, 4],\n     \"phases_skipped\": [6, 7],\n     \"total_pages_created\": 847,\n     \"total_entities_linked\": 1203,\n     \"next_phase\": 5\n   }\n   ```\n\n5. **Tell the user what to do next:**\n   > \"Your brain has N pages across people, calendar, email, and conversations.\n   > Live sync is configured for [sources]. From here:\n   > - The **signal-detector** captures entities from every conversation\n   > - The **briefing** skill can compile daily context\n   > - The **executive-assistant** pattern handles email triage\n   > - Say 'enrich [person]' to deep-dive any contact\"\n\n## Anti-Patterns\n\n- **Giving the agent raw OAuth tokens.** This is the #1 anti-pattern. An agent with\n  raw Gmail\u002FCalendar tokens is an uncontrolled attack surface — one prompt injection\n  and your entire Google account is exposed. Use ClawVisor. If the user declines\n  ClawVisor, skip to offline imports. Never offer direct OAuth as a fallback.\n- **Bulk importing everything without filtering.** The brain is for signal, not noise.\n  Filter out automated senders, marketing emails, utility conversations.\n- **Importing without entity cross-linking.** Every import should detect entities and\n  update existing brain pages. Isolated imports don't compound.\n- **Not gating on user consent.** Every phase should be presented as a choice. The user\n  may not want their DMs or therapy conversations imported.\n- **Importing everything at significance 1.** Not every conversation is worth a brain\n  page. Use the significance scale and skip utility content.\n- **Creating people pages for automated senders.** Sentry, GitHub notifications,\n  newsletter platforms are not people. Filter by the rules in Phase 4.\n\n## Resume Protocol\n\nIf the session is interrupted:\n\n1. Read `~\u002F.gbrain\u002Fcold-start-state.json`\n2. Skip completed phases\n3. Resume from `next_phase`\n4. The user doesn't have to repeat credential setup or re-import completed sources\n\n## Output Format\n\nAfter each phase:\n\n```\nPHASE N COMPLETE: [source name]\n================================\n\nPages created: N\nPages updated: N\nEntities linked: N\nTime elapsed: N min\n\nSample pages:\n- people\u002Fjane-smith.md (created — 3 emails, 5 meetings)\n- companies\u002Facme-corp.md (updated — 2 new employees linked)\n\nNext: Phase N+1 — [description]. Ready to proceed?\n```\n\n## Tools Used\n\n- `search` — check for existing pages before creating\n- `query` — hybrid search for entity deduplication\n- `get_page` — read existing pages for merge decisions\n- `put_page` — create and update brain pages\n- `add_link` — cross-reference entities\n- `add_timeline_entry` — record events on entity timelines\n- `sync_brain` — sync changes to the index after each phase\n",{"name":5,"tools":24,"version":32,"mutating":33,"triggers":34,"writes_to":44,"description":52,"writes_pages":33},[25,26,27,28,29,30,31],"search","query","get_page","put_page","add_link","add_timeline_entry","sync_brain","1.0.0",true,[35,36,37,38,39,40,41,42,43],"cold start","fill my brain","bootstrap brain","import my data","day one","get started","what should I import first","populate brain","now what?",[45,46,47,48,49,50,51],"people\u002F","companies\u002F","meetings\u002F","daily\u002F","media\u002F","conversations\u002F","sources\u002F","Day-one data bootstrapping for a new brain. Sequences the highest-leverage\ndata sources to go from empty brain to useful brain in one session. Uses\nClawVisor for safe credential handling — the agent never holds raw API keys.\nCovers Gmail import, calendar sync, contacts seeding, X\u002FTwitter archive,\nconversation imports, and file archives.\nUse when a user has just finished gbrain setup and asks \"now what?\"\n","main","6fd71a8e8b329f5375cf01a74a9da2af1909f4e0","skills\u002Fcold-start","https:\u002F\u002Fgithub.com\u002Fsinghvedant\u002Ftbrain","https:\u002F\u002Fgithub.com\u002Fsinghvedant","singhvedant1701.github.io","Web\u002FApp developer\r\nCloud Computing\r\nAI and ML","2026-07-04 14:54:38",{"source":8,"skillName":5,"commands":62,"prompt":72,"zipAvailable":33},[63,66,69],{"tool":64,"command":65},"npx","npx skills add singhvedant\u002Ftbrain --skill cold-start",{"tool":67,"command":68},"bunx","bunx skills add singhvedant\u002Ftbrain --skill cold-start",{"tool":70,"command":71},"pnpm","pnpm dlx skills add singhvedant\u002Ftbrain --skill cold-start","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fsinghvedant\u002Ftbrain\u002Fblob\u002Fmain\u002Fskills\u002Fcold-start\u002FSKILL.md",[],[75,81,87,94,101,107],{"skillNo":76,"name":77,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":78,"categories":79,"keywords":80,"publishedAt":19},"01M77MH99TG5CXS6ZC5V519J27","query-helper","https:\u002F\u002Fgithub.com\u002Fsinghvedant\u002Ftbrain\u002Fblob\u002Fmain\u002Ftest\u002Ffixtures\u002Fskill-catalog\u002Fskills\u002Fquery-helper\u002FSKILL.md",[],[],{"skillNo":82,"name":83,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":84,"categories":85,"keywords":86,"publishedAt":19},"01M77MH8A3A5PTQM0PQH6SCBA8","broken","https:\u002F\u002Fgithub.com\u002Fsinghvedant\u002Ftbrain\u002Fblob\u002Fmain\u002Ftest\u002Ffixtures\u002Fskill-catalog\u002Fskills\u002Fbroken\u002FSKILL.md",[],[],{"skillNo":88,"name":89,"nameZh":6,"description":90,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":91,"categories":92,"keywords":93,"publishedAt":19},"01M77MH6HHY2G5V0BXP8V316F8","brain-ops","Read, enrich, and write brain pages with source attribution.","https:\u002F\u002Fgithub.com\u002Fsinghvedant\u002Ftbrain\u002Fblob\u002Fmain\u002Ftest\u002Ffixtures\u002Fskill-catalog\u002Fskills\u002Fbrain-ops\u002FSKILL.md",[],[],{"skillNo":95,"name":96,"nameZh":6,"description":97,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":98,"categories":99,"keywords":100,"publishedAt":19},"01M77MH54BXVPFXK48DKBKFNRF","signal-detector","Always-on ambient signal capture for the OpenClaw reference fixture.","https:\u002F\u002Fgithub.com\u002Fsinghvedant\u002Ftbrain\u002Fblob\u002Fmain\u002Ftest\u002Ffixtures\u002Fopenclaw-reference-minimal\u002Fskills\u002Fsignal-detector\u002FSKILL.md",[],[],{"skillNo":102,"name":26,"nameZh":6,"description":103,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":104,"categories":105,"keywords":106,"publishedAt":19},"01M77MH3MPW52RQDQGTDR38MZT","Look up brain pages in the OpenClaw reference fixture.","https:\u002F\u002Fgithub.com\u002Fsinghvedant\u002Ftbrain\u002Fblob\u002Fmain\u002Ftest\u002Ffixtures\u002Fopenclaw-reference-minimal\u002Fskills\u002Fquery\u002FSKILL.md",[],[],{"skillNo":108,"name":109,"nameZh":6,"description":110,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":111,"categories":112,"keywords":113,"publishedAt":19},"01M77MH17X2DRV1DD6HV8Z89BA","context-now","ALWAYS-ON time-sensitivity discipline for the OpenClaw reference fixture.","https:\u002F\u002Fgithub.com\u002Fsinghvedant\u002Ftbrain\u002Fblob\u002Fmain\u002Ftest\u002Ffixtures\u002Fopenclaw-reference-minimal\u002Fskills\u002Fcontext-now\u002FSKILL.md",[],[]]