poc-intelligence
Generate a full POC Intelligence Report for a customer tenant as a self-contained HTML file. The report is customer-facing — clean language, no internal field names in the body, raw data in an Appendix.
When the agent picks this skill
Section titled “When the agent picks this skill”The user mentions any of: “POC report”, “CRM report”, “prepare for POC”,
“funnel analysis for <customer>”, “pipeline status for <customer>”,
“where should we focus for <customer>”, “what’s the CRM status for
<tenant>”, “analyze accounts for <customer>”, “help me win the POC”,
“conversion rate analysis”, “what titles are they targeting”, “contact
depth”, “inbound outbound split”, “plays for <customer>”, “how do we
move the numbers for <customer>”, “find detection engineers for
<customer>”, “map the SIEM for <customer>”.
What the skill teaches the agent
Section titled “What the skill teaches the agent”- Discover CRM structure dynamically. Never assume field names — every tenant’s Salesforce model is bespoke. The skill pulls metadata first.
- Synthesise across data sources. Salesforce (via the CRM integration proxy), Metabase, OnFire datasets (via
list_datasetsandquery_datasets), and Snowflake — combined into one consolidated report. - Report content:
- Three hero numbers (open pipeline ARR, win rate, closed lost ARR).
- Four data-grounded OnFire plays, each with 3 real account examples from Snowflake.
- Funnel health, competitive intelligence, deal velocity.
- Persona distribution and contact quality.
- Stalled deals + win/loss analysis.
- A POC playbook.
Underlying MCP tools
Section titled “Underlying MCP tools”get_tenant_settings ·
list_datasets ·
query_datasets ·
describe_dataset ·
search_people ·
ai_prospecting
External (outside this tool list): the CRM integration proxy,
Metabase (search, query, get_table, execute_query), and
Snowflake (sql_exec_tool).