sushidata-gtmgit n
A Claude Cowork plugin that turns Claude into a Sushidata-native GTM agent. Every research session, prospect list, and competitive finding is saved to the Sushidata context lake so future sessions build on what was already learned — not from scratch.
What this plugin does
Four overlapping use cases, all backed by the Sushidata context lake:
GTM Prospecting & Outbound — Discover ICP-fit companies with a Sushidata research swarm, find contacts via Hunter, verify emails, and activate campaigns in HeyReach. Sync everything to HubSpot. The full pipeline runs inside Cowork without switching tools.
Competitor & Market Intelligence — Research competitors channel by channel (SEO, paid, social, events, PR, hiring). Parse LinkedIn Ads JSON exports. Verify every claim and link before the document is shared. Save findings to the context lake for all future sessions.
ICP Signal Discovery — Upload Closed Won and Closed Lost domain lists. The niche-signal-discovery skill extracts multi-page website content and job listings, then computes Laplace-smoothed lift scores to surface what actually differentiates buyers from non-buyers.
Community & Campaign Intelligence — Ingest signals from Discord, Slack, forums, and campaign data into the context lake. Query them later to inform outreach, content, or product decisions.
Skills
| Skill | Trigger |
|---|
sushidata-research | Any research, prospecting, enrichment, or GTM execution task |
sushidata-sales-quickstart | "Show me what Sushidata can do", "find me some leads", "quick demo" |
sushidata-quickstart | "Research my competitors", "get started", first use with no prior context |
niche-signal-discovery | ICP signal analysis from won/lost domain lists |
Recipes
Step-by-step playbooks inside skills/sushidata-research/recipes/:
| Recipe | Use for |
|---|
gtm-competitor-report.md | Full GTM competitor analysis — channels, ads, events, PR, hiring |
document-accuracy-review.md | Verifying factual accuracy and link integrity in any research document |
account-orgchart.md | Mapping decision makers and warm intro paths for a target account |
build-tam.md | Building a total addressable market list from ICP criteria |
portfolio-prospecting.md | Finding companies backed by a specific investor, then contacts |
linkedin-url-lookup.md | Resolving LinkedIn profile URLs from names and companies |
scheduled-tasks.md | Setting up recurring or one-time automated GTM and research tasks |
small-business-prospecting.md | Finding local small businesses via Maps-style search |
Provider Playbooks
Inside skills/sushidata-research/provider-playbooks/:
| Playbook | Covers |
|---|
hunter.md | Domain search → email finder → email verifier chain |
heyreach.md | LinkedIn campaign insertion (≤50 contacts per batch) |
hubspot.md | Contact, company, note, and deal upsert patterns |
apify.md | Actor quality ranking, job listing scraping, LinkedIn data |
Sushidata API
All skills connect to:
BASE URL: https://dashboard.sushidata.ai/public/019e2876-d48b-765c-b303-06c7ff915c0d/
Tenant: Mitratech
Dataspace: Mitratech Internal
Key endpoints:
| Endpoint | Purpose |
|---|
POST /context/ | Save any finding to the context lake |
POST /query/ | Fast lookup — always try before deploying a swarm |
POST /swarm/deploy/ | Parallel research agents (swarmSize 2–20) |
POST /swarm/status/ | Poll swarm progress |
POST /verify/ | Validate evidence links before presenting to the user |
Validation Scripts
Two pure-Python (stdlib only) data quality scripts in skills/sushidata-research/scripts/:
validate-emails.py — Flags rows where the enriched email domain doesn't match the company domain. Warns if mismatch rate exceeds 20%.
python3 skills/sushidata-research/scripts/validate-emails.py enriched.csv \
--email-col email --domain-col domain
