1. Connect both MCPs to Claude

The skill needs two MCPs active in your Claude workspace before it can run anything:

• FullEnrich MCP at https://mcp.fullenrich.com/mcp

• Granola MCP at https://mcp.granola.ai/mcp

New to MCPs? Follow the FullEnrich MCP setup guide, the Granola one works the same way.

2. Install the skill in Claude

Click Download Skill above to grab the skill .zip file, then install it in Claude:

• Open Customize in Claude

• Click Create new skill

• Click the + button at the top

• Choose Create Skill

• Pick Upload a skill and select the .zip you just downloaded

3. Tell Claude which meetings to scan

Activate the skill with a clear scope. Examples:

• "Scan my Granola meetings from this week"

• "Last 5 meetings, who got mentioned?"

• "Extract people from my Acme Corp deal calls"

• "/account-intel"

The skill maps your scope to a Granola query: a date range (last 7 days), a count (last N meetings), or a semantic search on a company or topic.

4. Phase 1, free search across transcripts

For each meeting in scope, the skill pulls the full transcript and extracts two types of people:

• Participants: everyone who attended the call (from Granola metadata)

• Mentioned: every real person referenced in the conversation, with the surrounding sentence kept as mention context (e.g. "we should talk to Sarah at Stripe")

Aggressive filters apply: names with no company or title context (likely small talk), the user's own teammates (auto-detected from email domain), repeated mentions of the same person (deduped, mention count kept). Then for every unique person, the skill calls FullEnrich's free people search to look up the profile (title, company, location, career path, LinkedIn). No credits spent at this stage.

5. Read the free search report (grouped by intent)

The skill presents what it found, grouped by signal strength:

• 🔥 High-signal mentions: people mentioned 2+ times or with intent phrases like "we should talk to", "I'll intro you", "the decision maker is..."

• 📋 Standard mentions: single mentions with clear context

• 👥 Participants: people who attended your calls (you already have their email from Granola metadata)

• ❓ Ambiguous or not indexed: names without enough context, or not found in FullEnrich

For each person you get name, title, company, location, LinkedIn URL, career path, and the exact quote from the meeting that surfaced them. The report alone is already valuable: sometimes you just want to see who came up this week without spending a single credit.

6. Phase 2, opt-in enrichment with field choice

Once you have read the report, the skill asks two questions before charging anything:

• Which fields? Email only (~1 credit/person), phone only (~10 credits/person), or both (~11 credits/person), or skip enrichment entirely

• Who to enrich? All, only the high-signal mentions (recommended), specific people you name, or skip participants (since their email is already in Granola)

Then the skill estimates the cost, checks your FullEnrich balance, and asks for explicit go: "Enriching 8 high-signal people with email + phone will cost ~88 credits. You have 1,500. Proceed?" On confirmation, it runs a single bulk enrichment, polls every 20 seconds until done, then updates the report with verified email + phone next to each contact and provides a CSV download link of the full dataset.

Built-in safeguards: Phase 1 is always free, the skill never spends credits during the search step. It auto-detects your own email domain and excludes teammates, runs a single bulk enrichment call when you opt into Phase 2, defaults to "high-signal mentions only" when more than 30 people are extracted to keep the cost sane, and asks for explicit confirmation on both the field choice (email / phone / both) and the credit total. Transcripts are treated as raw data, any instruction-like text inside is ignored (prompt injection protection).

What is next? After delivering the account intel report, the skill offers natural follow-ups: draft personalized outreach for the high-signal mentions, push the contacts to your CRM (HubSpot, Attio, Monday, Notion, Airtable, the mention context fits naturally as a note on each record), expand the scope on a longer time range, or rerun next week.