Person Dossier

Build or refresh a dossier when relationship research is requested. An incidental mention is not a dossier request. Use only relevant authorized sources.

Trigger

Relevant requests include:

• strategy or planning conversations
• follow-up decisions
• relationship context questions
• opportunity/deal/property discussions

Output Path

• Vault/Resources/<Person Name> - Dossier.md

Required Sections

1. All Email Threads
2. Key Topics Discussed
3. Projects Mentioned
4. Properties or Deals Referenced
5. Timeline of Interactions
6. Current Status of Open Items

Workflow

1. Validate runtime and auth

secrets status
secrets lease front_api_token --ttl 15m

If secrets service is down:

secrets serve &

2. Resolve identifiers

Collect:

• primary name
• known email(s)
• aliases/handles

Use these as query inputs for all sources.

3. Pull Front threads (primary)

joelclaw email inbox -q "<person name>" -n 200
joelclaw email inbox -q "from:<email-or-handle>" -n 200

Read each matching conversation:

joelclaw email read --id <cnv_id>

4. Pull Granola meetings

List meetings in a practical window:

granola meetings --range this_week
granola meetings --range this_month

For each candidate meeting ID:

Use recall and the current session-search contract. Keep private queries in structured MCP arguments. Raw evidence requires a scope-bound evidenceDrilldownReceipt; missing projections do not grant raw access.

5. Pull Brain-backed recall context

Use recall and the current session-search contract. Keep private queries in structured MCP arguments. Raw evidence requires a scope-bound evidenceDrilldownReceipt; missing projections do not grant raw access.

6. Pull event timeline context

joelclaw events --hours 720 --count 200
joelclaw events --prefix meeting --hours 720 --count 200

7. Normalize extracted data

Populate these buckets from evidence:

• email threads: subject, conversation id, participants, last activity date
• topics: recurring discussion themes
• projects: named initiatives
• properties/deals: assets, opportunities, transactions
• timeline: dated interaction entries by channel
• open items: owner, status, due date, next step

8. Create or update dossier markdown

If file does not exist: create from template.

If file exists: update in-place.

Update rules:

• append new timeline entries without removing older ones
• dedupe by stable key (channel + source_id + date + summary)
• merge open item status updates instead of creating duplicates
• keep Verified and Inferred labels on each material claim
• refresh Last Updated timestamp

9. Record evidence and confidence

Each section must indicate provenance:

• Verified from Front thread
• Verified from Granola meeting
• Inferred from Brain-backed recall/event context

Never present inferred content as verified.

10. Blocker handling

If blocked:

1. record exact technical error text
2. try authorized event metadata or scoped recall; do not bypass transcript access requirements
3. produce a blocked dossier with explicit gaps
4. include exact next-run commands

No silent partials.

Dossier Template

# <Person Name> - Dossier

## Identity
- Name: <name>
- Emails: <email1>, <email2>
- Aliases: <alias1>, <alias2>
- Last Updated: YYYY-MM-DD

## 1. All Email Threads with <Person Name>
- Thread: <subject>
  - Conversation ID: <cnv_id>
  - Last activity: YYYY-MM-DD
  - Evidence: Verified from Front thread

## 2. Key Topics Discussed
- <topic> (Verified/Inferred)

## 3. Projects Mentioned
- <project> (Verified/Inferred)

## 4. Properties or Deals Referenced
- <item> (Verified/Inferred)

## 5. Timeline of Interactions
- YYYY-MM-DD | <channel> | <summary> | <source_id>

## 6. Current Status of Open Items
- Item: <description>
  - Owner: system owner | other person | unknown
  - Status: open | blocked | closed
  - Next step: <text>
  - Due: YYYY-MM-DD | unknown
  - Evidence: <source>

## Evidence Index
- Front conversation IDs: <list>
- Granola meeting IDs: <list>
- Memory queries: <list>
- Event windows: <list>

## Blockers
- <none or error text>

Quick Build Command Set

Use recall and the current session-search contract. Keep private queries in structured MCP arguments. Raw evidence requires a scope-bound evidenceDrilldownReceipt; missing projections do not grant raw access.

Quality Bar

• No fabricated facts
• Open items always include owner
• Timeline uses concrete dates whenever available
• Verified and inferred content clearly separated