Blitz Preferences
Session preferences that control how AI handles emails during blitz sessions. Set once, applied every session.
Purpose
Blitz preferences store persistent settings for how you want email handled: your sign-off style, default tone, noise handling, and delegation rules. The AI checks these before every session.
Preferences Structure
{ "newsletters": "archive", "noise_handling": "auto_archive", "default_tone": "professional", "sign_off": "Cheers,\nStuart", "delegation": [ { "match": "billing|invoice|payment", "cc_email": "[email protected]", "reason": "Financial matters" } ], "priorities": { "domains": ["bigclient.com", "investor.co"], "keywords": ["urgent", "asap", "deadline"], "senders": [] }, "quick_responses": { "thanks_acknowledged": "Thanks for this - acknowledged and noted.", "will_review": "Thanks, I'll review and get back to you shortly.", "meeting_confirmed": "Confirmed - see you then." }, "setup_complete": true}Field Reference
| Field | Values | Description |
|---|---|---|
newsletters | archive, unsubscribe, ask | How to handle newsletter emails |
noise_handling | auto_archive, ask | Receipts, shipping notifications, etc. |
default_tone | casual, professional, formal | Default response tone (can be overridden per contact) |
sign_off | String with \n | How to sign emails (e.g., "Cheers,\nStuart") |
delegation | Array of rules | When to CC others (match pattern, email, reason) |
priorities | Object | Domains, keywords, senders that get high priority |
quick_responses | Object | Short phrases for common quick replies |
setup_complete | Boolean | If false, AI asks setup questions on first run |
First Run Setup
If setup_complete is false or the file is empty, the AI will ask:
- "When I see newsletters, should I archive them, offer to unsubscribe, or ask you each time?"
- "How do you like to sign off emails?"
- "Is there anyone I should CC for specific types of issues?"
- "For obvious noise (receipts, shipping notifications), should I auto-archive or ask first?"
- "What's your default tone? (casual, professional, formal)"
Delegation Rules
Delegation rules tell the AI when to CC someone on responses:
"delegation": [ { "match": "billing|invoice|payment", "cc_email": "[email protected]", "reason": "Financial matters" }, { "match": "support|help|issue|bug", "cc_email": "[email protected]", "reason": "Customer support" }, { "match": "contract|agreement|legal", "cc_email": "[email protected]", "reason": "Legal matters" }]The match field is a regex pattern. When an email matches, the AI will suggest (or automatically add in auto mode) the CC.
Quick Responses
Short phrases for common quick replies:
"quick_responses": { "thanks_acknowledged": "Thanks for this - acknowledged and noted.", "will_review": "Thanks, I'll review and get back to you shortly.", "meeting_confirmed": "Confirmed - see you then.", "received_will_process": "Got it - I'll process this and follow up."}You can reference these in sessions: "Send a thanks_acknowledged reply."
Priority Settings
Define what counts as high priority:
"priorities": { "domains": ["bigclient.com", "investor.co", "partner.io"], "keywords": ["urgent", "asap", "deadline", "critical"], "senders": ["[email protected]"]}Emails matching these are flagged for immediate attention and never auto-archived.
File Location
~/.config/pontius/knowledge/blitz/preferences.jsonHow AI Uses Preferences
- Check on every blitz session — Load preferences before handling emails
- Apply consistently — Use sign_off, tone, delegation rules
- Respect noise settings — Don't ask about newsletters if user said auto-archive
- Update when directed — If user changes preference mid-session, update the file
Ending a Session: Knowledge Capture
When you've cleared your inbox, run pontius zero to capture what was learned:
$ pontius zeroThis prompts your AI to reflect on the session and update the knowledge base with:
- New contacts — People you interacted with who should be remembered
- Patterns — Recurring email types that could be handled automatically next time
- Templates — Replies you drafted that might be useful again
- Corrections — Anything that was wrong or outdated in existing knowledge
- Policies — Preferences you expressed about how to handle things
This creates a natural learning loop — each blitz session makes the next one smoother.
pontius zero at the end of every blitz. Even if there's nothing new to capture, it takes seconds and ensures your knowledge base stays current.