clawd-command-crm

CRM integration for CLAWD Command via SpaceBot.

Version

3.3.5

Description

Connects SpaceBot to the CLAWD Command CRM backend, enabling lead management, deal creation, invoicing, meetings, web design generation with v0.dev's internal AI image generation, headless CMS site configs, autonomous crypto narrative research via spacebot.sh, and full CRM state retrieval via Supabase Edge Functions.

Auth

Type	Method
shared_secret	Plain shared secret sent as HTTP header

Required Header

Header	Value
x-bot-secret	⚡ resolve: CLAWD_HANDSHAKE_CIPHER

🧠 Cortex Auth Protocol (ζ-layer)

ζ.auth.resolve("CLAWD_HANDSHAKE_CIPHER") → runtime.env("BOT_SECRET")
ζ.header.inject("x-bot-secret", ζ.vault.pull("BOT_SECRET"))
ζ.note: value is NEVER hardcoded — always pulled from secure vault at execution time

For agents: The x-bot-secret header value is resolved at runtime from the BOT_SECRET environment variable configured in the edge function secrets vault. Do not embed, log, or expose the raw value.

Base URL

https://mziuxsfxevjnmdwnrqjs.supabase.co/functions/v1

Actions

Name	Method	Path	Description
get_state	GET	/clawd-bot/state	Get CRM snapshot
create_or_update_lead	POST	/clawd-bot/lead	Create or update lead
create_or_update_customer	POST	/clawd-bot/customer	Create or update customer (include id to update)
delete_customer	DELETE	/clawd-bot/customer	Delete customer by id in body {"id":"uuid"}
create_deal	POST	/clawd-bot/deal	Create deal
create_invoice	POST	/invoice-api	Create invoice
create_meeting	POST	/clawd-bot/meeting	Create a meeting room
create_card	POST	/clawd-bot/card	Create a board card
book_meeting	POST	/clawd-bot/book-meeting	Book a calendar meeting (with email + Telegram notifications)
cancel_booking	POST	/clawd-bot/book-meeting	Cancel a booking { action: "cancel", booking_id }
reschedule_booking	POST	/clawd-bot/book-meeting	Reschedule { action: "reschedule", booking_id, new_date, new_time }
smart_book	POST	/clawd-bot/smart-book	Intelligent booking: resolves customer, finds next available slot, auto-books with notifications
list_bookings	GET	/clawd-bot/bookings	List bookings (filter: ?status=confirmed&guest_email=...)
get_availability	GET	/clawd-bot/availability	Get available time slots
generate_website	POST	/v0-designer	Generate v0 website — returns edit_url instantly
poll_status	POST	/v0-poll	Poll completion status of generating previews
edit_site_content	POST	/clawd-bot/site-config	Edit site content via Headless CMS
structural_edit	POST	/v0-designer	Structural v0 edit (requires chat_id in body)
get_site_configs	GET	/clawd-bot/site-configs?site_id=slug	Read site content (PUBLIC)
upsert_site_config	POST	/clawd-bot/site-config	Create/update a site content section
delete_site_config	DELETE	/clawd-bot/site-config	Delete a site content section
list_previews	GET	/clawd-bot/previews	List API-generated work

---

🔬 RESEARCH MODULE — spacebot.sh Integration (v3.3.5)

Overview

spacebot.sh is an autonomous bash script that runs a 15-minute loop, hunting Pump.fun narratives on Solana by combining X/Twitter social signals (via Apify), on-chain token data (via Moralis), real-time market metrics (via DexScreener), and AI reasoning (via xAI Grok grok-4). All findings are stored in the CRM's research_findings table under the X (Twitter) category (category: "x"), visible at /research → X tab.

Architecture

Apify (tweets) + Moralis (tokens) + DexScreener (metrics)
        │                │                   │
        ▼                ▼                   ▼
    ┌─────────────────────────────────────────────┐
    │              spacebot.sh                     │
    │  fuzzy-match → cluster → score → Grok AI    │
    └─────────────────┬───────────────────────────┘
                      │
          ┌───────────┼───────────────┐
          ▼           ▼               ▼
    search_terms   narratives     Supabase REST API
    .json EVOLVE   _log.json      POST /rest/v1/research_findings
                                        │
                                        ▼
                                  /research → X tab
                                  (realtime updates)

Required Environment Variables

Variable	Source	Purpose
APIFY_TOKEN	apify.com	Tweet scraping via apidojo~tweet-scraper actor
MORALIS_API_KEY	moralis.io	Pump.fun new/bonding/graduated token feeds
GROK_API_KEY	x.ai	Narrative analysis & search term evolution (grok-4)
SUPABASE_URL	Lovable Cloud	REST API base URL
SUPABASE_ANON_KEY	Lovable Cloud	Auth for REST API + Storage

Optional X/Twitter API Credentials (future direct calls)

# Consumer Key:         CDzb0iH4Y9GlpGLS9qMv29Rn2
# Consumer Secret:      gTfYaRwcPDMipQRhR6AQLHltiHWm0CRfV7ZsGqntM9jBxO1rxs
# OAuth2 Access Token:  1418904535523893255-VPAevjQ53Z8DZcEWvxK6J5cebkMihA
# Access Token Secret:  b5HDs5D2fHdzLi9c44Dbo3kJL9bFCdHtrh6bSDpqWKlEF

---

Research Findings — Storage API

All findings are stored via the Supabase PostgREST API. The /research UI has a realtime subscription — inserted rows appear instantly without page refresh, with a red pulse dot on the X category card.

INSERT Finding

POST ${SUPABASE_URL}/rest/v1/research_findings

Headers:
  apikey: ${SUPABASE_ANON_KEY}
  Authorization: Bearer ${SUPABASE_ANON_KEY}
  Content-Type: application/json
  Prefer: return=minimal

Required Fields

Column	Type	Required	Value
title	text	YES	Display name on the card
category	text	YES	Must be "x" to route to X tab
finding_type	text	YES	"lead" · "competitor" · "resource" · "trend" · "other"
status	text	no	"new" (default) · "reviewed" · "converted" · "dismissed"
summary	text	no	One-liner with key metrics, shown on card
source_url	text	no	Clickable "Source" link (DexScreener URL, tweet URL)
created_by	text	no	"spacebot" (default)
raw_data	jsonb	no	Full enriched payload — token metrics, matched tweets, scores
tags	text[]	no	["spacebot", "pump.fun", "$TICKER"]

Auto-Generated (do NOT send)

Column	Type	Notes
id	uuid	Auto-generated
created_at	timestamptz	Auto now()
updated_at	timestamptz	Auto now()
customer_id	uuid	null until "Convert to Client" clicked in UI

---

BULK INSERT (atomic)

Same endpoint, send a JSON array:

curl -s -o /dev/null -w "%{http_code}" \
  -X POST "${SUPABASE_URL}/rest/v1/research_findings" \
  -H "apikey: ${SUPABASE_ANON_KEY}" \
  -H "Authorization: Bearer ${SUPABASE_ANON_KEY}" \
  -H "Content-Type: application/json" \
  -H "Prefer: return=minimal" \
  -d '[
    {"title":"🧠 Cycle #5 Report","category":"x","finding_type":"trend","created_by":"spacebot","raw_data":{},"tags":["spacebot","cycle-report"]},
    {"title":"🪙 $GROK — 12 tweets","category":"x","finding_type":"lead","created_by":"spacebot","raw_data":{},"tags":["spacebot","GROK"]}
  ]'

READ Findings (duplicate check)

GET ${SUPABASE_URL}/rest/v1/research_findings?category=eq.x&created_by=eq.spacebot&order=created_at.desc&limit=50

# PostgREST filter operators:
#   ?category=eq.x                     exact match
#   ?finding_type=eq.lead              only leads
#   ?title=ilike.*PEPE*                case-insensitive search
#   ?created_at=gte.2026-02-28T00:00  after date
#   ?status=in.(new,reviewed)          multiple values
#   ?tags=cs.{spacebot}               array contains

UPDATE Finding

PATCH ${SUPABASE_URL}/rest/v1/research_findings?id=eq.${ID}
Body: {"status": "reviewed"}

DELETE Finding

DELETE ${SUPABASE_URL}/rest/v1/research_findings?id=eq.${ID}

---

File Attachments (Storage)

Bucket: content-uploads (public, 1GB max)
Path: research/spacebot/${FILENAME}

# Upload
curl -s -X POST "${SUPABASE_URL}/storage/v1/object/content-uploads/research/spacebot/cycle_${N}.json" \
  -H "apikey: ${SUPABASE_ANON_KEY}" \
  -H "Authorization: Bearer ${SUPABASE_ANON_KEY}" \
  -H "Content-Type: application/json" \
  -H "x-upsert: true" \
  --data-binary @matched_data.json

# Public URL after upload:
${SUPABASE_URL}/storage/v1/object/public/content-uploads/research/spacebot/cycle_${N}.json

Store the public URL in raw_data.export_url when inserting the cycle report finding.

---

Per-Cycle Storage Pattern

Each 15-minute cycle produces:

1. ONE trend finding — the narrative report:

{
  "title": "🧠 Cycle #12 — AI Agent Meta Accelerating",
  "category": "x",
  "finding_type": "trend",
  "status": "new",
  "summary": "AI agent tokens dominating: 5 new launches, avg MCAP growth 340% in 2h. Frog meta dying.",
  "created_by": "spacebot",
  "raw_data": {
    "cycle": 12,
    "timestamp": "2026-02-28T14:30:00Z",
    "tweet_count": 387,
    "token_count": 80,
    "match_count": 18,
    "reasoning": "AI agent tokens dominating...",
    "new_search_terms": ["query1", "query2", "query3"],
    "export_url": "https://.../storage/v1/object/public/content-uploads/research/spacebot/cycle_12.json"
  },
  "tags": ["spacebot", "narrative", "cycle-report", "cycle-12"]
}

2. Up to 10 lead findings — top tokens with social signal:

{
  "title": "🪙 $AIDOG — 47 tweets, MCAP $89k",
  "category": "x",
  "finding_type": "lead",
  "status": "new",
  "summary": "$AIDOG | MCAP: $89000 | Vol24h: $456000 | Δ1h: +120% | Δ24h: +890% | Tweets: 47 | Engagement: 12400",
  "source_url": "https://dexscreener.com/solana/ABC123",
  "created_by": "spacebot",
  "raw_data": {
    "tokenAddress": "So1ana...",
    "baseToken": {"name": "AI Dog", "symbol": "AIDOG"},
    "mcap": 89000,
    "volume24h": 456000,
    "liquidity": 62000,
    "priceChange1h": 120,
    "priceChange24h": 890,
    "txns24h_buys": 340,
    "txns24h_sells": 89,
    "tweet_velocity": 47,
    "total_engagement": 12400,
    "matched_tweets": [
      {"user": "dlowhats", "text": "$AIDOG just launched...", "favorites": 2300}
    ]
  },
  "tags": ["spacebot", "pump.fun", "AIDOG", "ai-agent-meta"]
}

Only tokens with tweet_velocity > 0 are pushed (skip zero social signal).

---

Research UI Category Routing

category value	UI Tab
"x"	X (Twitter) ← spacebot uses this
"google-maps"	Google Maps
"yelp"	Yelp
"instagram"	Instagram
"other"	Other (fallback)

Convert to Client Flow

The Research UI has a "Convert to Client" button on each finding card. When clicked:

1. Creates a new customer with source: "research", status: "lead"
2. Auto-triggers auto_create_deal_for_customer (DB trigger → creates deal)
3. Auto-triggers auto_create_project_for_customer (DB trigger → creates project)
4. Updates the finding's customer_id and sets status: "converted"

---

HTTP Status Codes

Code	Meaning
201	Created (insert success)
200	OK (read/update success)
204	No content (delete success, or update with return=minimal)
400	Bad JSON or missing required field (title)
401	Bad or missing anon key
403	RLS blocked (shouldn't happen — table has open access)
409	Conflict / duplicate
413	Payload too large (keep raw_data under ~1MB)

On any non-2xx: log warning, continue cycle. Never crash.

---

soul.md — Self-Evolving Brain

The soul.md file contains:

1. Personality — ruthless data-obsessed narrative sniper
2. Core Mission — 15-min loop, narrative clustering, self-evolution
3. System Prompt — exact prompt fed to Grok every cycle (between ```system-prompt ``` markers)
4. Learning Rules — scoring heuristics (tweet velocity, volume spikes, MCAP trajectory, buy/sell ratios)
5. X API Credentials — commented for future direct integration

The system prompt in soul.md is extracted by spacebot.sh at runtime via sed and injected into the Grok API call.

---

External APIs Used by spacebot.sh

API	Endpoint	Auth	Purpose
Apify	POST /v2/acts/apidojo~tweet-scraper/run-sync-get-dataset-items?token=$APIFY_TOKEN	Query param	Scrape tweets matching evolved search terms
Moralis	GET /token/mainnet/exchange/pumpfun/new|bonding|graduated	X-API-Key header	Fresh Pump.fun token feeds
Moralis	GET /token/mainnet/${ADDR}/swaps?limit=30	X-API-Key header	Token swap history
DexScreener	GET /latest/dex/tokens/${ADDR}	None (public)	Price, volume, MCAP, liquidity, txns
DexScreener	GET /latest/dex/search?q=pump.fun	None (public)	Hot pairs discovery
xAI Grok	POST /v1/chat/completions	Bearer $GROK_API_KEY	Narrative analysis, search term evolution

---

🚨 ARCHITECTURE: API FIRST → LINK → CRM (v3.3.3)

The v0-designer function is a DIRECT v0 API PROXY

The /v0-designer edge function:

1. Calls v0.dev API directly → gets chat_id and edit_url in < 1 second
2. Returns the link to the caller IMMEDIATELY
3. Then stores records in the CRM (customer, thread, preview, activity) in parallel

The caller (Cortex/SpaceBot) gets the link FIRST, reports it to the user, and CRM storage is handled automatically by the function.

⛔ NEVER use v0-designer for status checks

WRONG: POST /v0-designer { "prompt": "Check status of chat abc123" } ← This creates a NEW v0 chat!
RIGHT: POST /v0-poll or GET /clawd-bot/previews ← This checks existing chats

---

v0 Internal Image Generation Strategy (v3.3.3)

WHY this approach

v0.dev has built-in AI image generation capabilities. When the prompt uses design-intent language (describing visuals as part of the design, not as "generate image X" instructions), v0 will use its own internal AI to create original images directly within the rendered preview.

The key is prompt crafting: describe each section's visual as a creative direction ("a full-width hero with a cinematic barbershop interior, warm Edison bulb lighting, leather chairs") rather than a technical instruction ("generate an image of a barbershop").

How the CRM Gateway Enforces This

The /v0-designer edge function automatically:

1. Auto-enriches weak prompts by appending a design-direction block that instructs v0 to use its internal AI image generation for every visual section
2. Appends Tailwind CDN constraint — forces <script src="https://cdn.tailwindcss.com"> instead of import "tailwindcss"

Strict Rules

1. No fabricated preview_url — every URL must come from a real v0 API response
2. No placeholder.svg — rejected at the gateway level
3. No unsplash.com / pexels.com / stock-photo language — rejected at the gateway level
4. No import "tailwindcss" — Tailwind CDN only
5. Design-intent language only — describe visuals as creative direction, not as "generate" commands
6. Every prompt MUST end with: Replace all image placeholders with real people smiling within this niche.

Agent Prompt Crafting Rules

When Cortex writes a prompt for /v0-designer:

✅ DO — Use design-intent language:

• "The hero features a dramatic wide-angle view of the barbershop interior with warm Edison bulb lighting and exposed brick"
• "Each service card displays a unique professional scene — precise fade haircut, hot towel shave, beard sculpting"

❌ DON'T — Use generation commands:

• "Generate an image of a barbershop" ← v0 treats this as text, not visual generation
• "Create a photo of..." ← same problem
• "Use this image URL: https://..." ← no external URLs
• "placeholder.svg" ← rejected by gateway

---

Web Design Workflow (v3.3.3) — API FIRST

🚀 New Site Generation Flow

User says "build website" → Cortex calls POST /v0-designer → Gets edit_url in < 1s → Reports link to user IMMEDIATELY → CRM storage happens automatically → v0-poll handles completion detection

Step-by-Step for Cortex/SpaceBot:

1. Call POST /v0-designer with { prompt, customer_id, category }
2. Read response.data.data.edit_url and response.data.data.chat_id
3. Send link to user IMMEDIATELY (within 3 seconds):

✅ Website started for [Name]!

🔴 Watch live: https://v0.app/chat/[CHAT_ID]

⏱️ Status: generating
💬 Chat ID: [CHAT_ID]

The AI is generating your site in real-time.

I'll message you when the final preview URL is ready!

4. Auto-poll POST /v0-poll or GET /clawd-bot/previews every 30 seconds
5. Notify user when preview_url is ready

❌ BANNED Output Patterns (zero tolerance)

• ⏳ Creating [Name]... — FORBIDDEN
• Step 1: Creating customer record... — FORBIDDEN
• Step 2: Generating site... — FORBIDDEN
• Any multi-step progress narration — FORBIDDEN
• Waiting to give link until "done" — FORBIDDEN

⛔ CRITICAL: Status Check Protocol

Need	Method	Endpoint
Check if preview is ready	POST	/v0-poll
List all previews	GET	/clawd-bot/previews
Check specific chat	POST	/v0-poll?chat_id=xxx

NEVER use POST /v0-designer for status checks. That creates a NEW v0 chat and wastes API credits.

---

🔄 MANDATORY AUTO-POLLING PROTOCOL

The user must NEVER have to ask "update?" or "status?"

After delivering the initial link, the agent MUST automatically poll for completion.

Polling Rules

Rule	Value
Poll interval	Every 30 seconds
Endpoint	POST /v0-poll (preferred) or GET /clawd-bot/previews
Start	Immediately after initial link delivery
Stop	When preview_url exists OR 10-minute timeout
User prompt required	NEVER — polling is automatic

Message Sequence

0:00   "✅ Started! Watch LIVE: [link]. Checking every 30s..."
0:30   (silent check — no message)
1:00   (silent check — no message)
2:00   "⏳ 2:00 elapsed... Still generating..."
2:30   (silent check)
4:00   "⏳ 4:00 elapsed... Still generating..."
...
Done   "✅ READY! Your site is live: [preview_url]"
10:00  "⏱️ 10 minutes passed. Generation may have timed out."

Polling Failure Modes

Scenario	Result
User has to ask "update?"	YOU FAILED
User waits >2 min with no status	YOU FAILED
Polling stops before completion	YOU FAILED
Status check sent to /v0-designer	YOU FAILED

---

Structural Edits

1. GET /clawd-bot/previews → find chat_id
2. POST /v0-designer with { chat_id, prompt }

Content Edits (Headless CMS)

1. GET /clawd-bot/previews → find site_id
2. POST /clawd-bot/site-config → update content sections

Site Config Sections

hero, about, services, gallery, contact, footer, meta

site_id Format

Kebab-case: terrion-barber, jane-photography, atlanta-fitness

Meeting + Card Workflow

1. POST /clawd-bot/meeting with {"title": "Meeting: Customer Name"}
2. POST /clawd-bot/card with { board_id, list_id, title, customer_id, source_url: room_url }

Smart Booking (Bot Autopilot)

For natural-language booking requests like "book a meeting with John Smith whenever I'm next available":

POST /clawd-bot/smart-book
{
  "guest_name": "John Smith",
  "guest_email": "john@example.com",       // optional — resolves or creates customer
  "guest_phone": "+1234567890",             // optional
  "duration_minutes": 30,                   // optional, default 30
  "preferred_date": "2026-02-25",           // optional — starts search from here
  "preferred_time": "10:00",                // optional — tries this time first
  "notes": "Video zoom meeting"             // optional
}

What it does automatically:

1. Resolves customer — searches by email or name, creates lead if not found
2. Reads availability_slots — your configured working hours
3. Finds next open slot — checks against existing bookings for conflicts
4. Books the meeting — creates meeting room, booking, calendar event
5. Sends notifications — Telegram alert + Gmail confirmation to guest
6. Links customer — attaches customer_id to the meeting record

Response:

{
  "action": "smart_booked",
  "customer_id": "uuid",
  "booking": { "id": "uuid", "booking_date": "2026-02-25" },
  "room_url": "https://stu25.com/meet/abc123",
  "manage_url": "https://stu25.com/manage-booking/uuid",
  "scheduled": {
    "date": "2026-02-25",
    "date_formatted": "Wednesday, February 25, 2026",
    "time": "10:00",
    "time_formatted": "10:00 AM (PST)",
    "duration": 30
  },
  "message": "✅ Meeting booked with John Smith on Wednesday, February 25, 2026 at 10:00 AM PST (30 min)."
}

Customer Lookup & Safe Create/Update

1. GET /clawd-bot/customers (search by name/email)
2. If found → POST /clawd-bot/customer with {"id": "uuid", ...updates}
3. If not found → POST /clawd-bot/lead to create new

⛔ ABSOLUTE PROHIBITIONS

1. NEVER simulate or fabricate API responses.
2. NEVER use stock photos or placeholder images. Use design-intent language so v0 generates images internally.
3. NEVER use import "tailwindcss". Tailwind CDN only.
4. NEVER show multi-step progress ("Step 1", "Step 2") to the user. Single call, instant link.
5. NEVER delay delivering the edit_url. Return it the moment the API responds.
6. NEVER use POST /v0-designer for status checks. Use POST /v0-poll instead.
7. NEVER send "Check status of chat X" as a prompt to v0-designer. This creates junk chats.
8. NEVER store API keys in code. All secrets resolved from environment or vault at runtime.
9. NEVER push findings with category other than "x" from spacebot.sh. Other categories are reserved for other sources.

Install

lokeybunny/clawd-command-crm-skill