arcascience-email（阿卡科学电子邮件）

ArcaScience Email Assistant

Description

Comprehensive email management assistant for Romain at ArcaScience. Reads inbox, triages emails, drafts contextual replies using pharma intelligence dossiers, and manages outreach. Adapts tone, language, and content based on the recipient's profile and relationship.

Trigger

Use this skill when the user asks to:

• Read, check, scan, or list emails
• Draft, write, compose, or reply to emails
• Handle unanswered or pending emails
• Prepare outreach to pharma contacts
• Send a follow-up or thank-you email
• Manage inbox or email workflow

Also triggered automatically by the daily-email-drafts cron job.

Sender Identity

• Name: Romain
• Email: romain@arcascience.ai
• Role: Business development / strategic partnerships at ArcaScience
• Company: ArcaScience — AI-powered Benefit-Risk Assessment platform for pharma

Instructions

Step 1: Fetch Emails

Use the Gmail skill (Maton API) to fetch emails from romain@arcascience.ai.

• For "check my emails" → fetch last 10-20 unread or recent
• For "emails from today" → fetch last 24 hours
• For "reply to X" → fetch the specific thread
• For daily cron → fetch last 24 hours, all emails

Read the FULL content of each email (not just subject/snippet).

Step 2: Triage

Classify every email into one of these categories:

Present the triage as a summary table FIRST, then drafts for each ✅ email.

Step 3: Identify Sender Context

For each email requiring a reply:

A. Check if sender is internal:

• @arcascience.ai emails → internal colleague
• Known internals: Julie Green (julie@arcascience.ai), María Lola Álvarez (marialola@arcascience.ai)

B. Check if sender is from a known pharma company:

• Look up arcascience_dossiers/companies/ for company dossier
• Look up arcascience_dossiers/stakeholders/ for contact profile
• Check arcascience_dossiers/master_targeting_table.csv for targeting data
• If found: extract strategic entry angle, key challenges, therapeutic focus, buying pathway

C. If sender is unknown external:

• Infer company and role from email signature/domain
• Search web if needed for company context
• Draft a professional but measured response

Step 4: Draft Replies

Language Rule (ABSOLUTE — NO EXCEPTIONS)

Reply in the SAME language as the incoming email. If email is in French → draft in French. If in English → English. If in German → German. Never translate.

Tone Adaptation Matrix

Draft Structure by Email Type

For pharma business emails (prospects/clients):

1. Opening: Acknowledge their message or context
2. Value bridge: Connect ArcaScience BRA platform to THEIR specific challenge (use dossier intelligence)
3. Proof point: Reference relevant capability, metric, or client base
4. Call to action: Suggest a concrete next step
5. Sign-off: Professional

For internal emails:

1. Address the question/task directly
2. Provide information or confirm action
3. Next steps if needed
4. Short sign-off

For admin/operational emails:

1. Confirm receipt or action
2. Provide requested information
3. Done

Step 5: Present Drafts

For each draft, present in this format:

────────────────────────────────────────
📧 DRAFT [N] — [Context description]
To: [recipient]
Subject: Re: [subject]
Langue: [detected language]
Ton: [tone level]
Dossier: [company dossier found / not found]
────────────────────────────────────────

[Draft body]

────────────────────────────────────────
✅ Dis "envoyer [N]" pour envoyer
✏️ Dis "modifier [N]: [instructions]" pour ajuster
────────────────────────────────────────

Step 6: Attach Documents When Relevant

Before sending, evaluate if the email would benefit from a piece jointe (PJ). Proactively suggest attachments — don't wait for Romain to ask.

When to Attach

Attachment Source Directories

Sales Package (original PDFs — use these for email attachments): C:\Users\Romai\Downloads\Sales Package\

Client-specific decks (markdown exports in the repo): C:\Users\Romai\OneDrive\Documents\GitHub\Decks-Docs\docs\

• Sanofi decks, Servier decks, Hyloris workshops, EDSC roundtable, MHRA slides

How to Attach

When drafting, if an attachment is relevant:

1. Add a PJ: line to the draft presentation:

────────────────────────────────────────
📧 DRAFT [N] — [Context]
To: [recipient]
Subject: Re: [subject]
Langue: [detected language]
Ton: [tone level]
Dossier: [company dossier found / not found]
📎 PJ: 2025_01_Arcascience_OnePager_VF.pdf
────────────────────────────────────────

1. When sending via Gmail API, attach the file from the Sales Package path
2. If multiple attachments are relevant, list them all but limit to max 3 per email (keep it focused)
3. ALWAYS mention the attachment in the email body (e.g., "Vous trouverez en PJ notre one-pager..." or "Please find attached...")

Attachment Decision Rules

• First contact → ALWAYS attach the general one-pager
• Prospect mentions a specific therapeutic area → Attach clinical landscape one-pager
• Post-meeting follow-up → Attach whatever was discussed (proposal, deck, etc.)
• Legal/partnership discussion → Attach CDA or partnership contract blank
• Never attach contracts to cold prospects or unknown contacts
• Never attach internal-only documents (Sales Organisation package) to external contacts
• When in doubt → suggest the attachment to Romain in the draft and let him confirm

Step 7: Send (ONLY on explicit confirmation)

CRITICAL: NEVER send an email without Romain saying "envoyer X" or "send X" explicitly.

When confirmed:

1. Send via Gmail skill with attachment(s) if specified
2. Confirm delivery and list of PJ sent
3. Log the sent email

ArcaScience Quick Reference (for drafting)

What we do: AI-powered dynamic Benefit-Risk Assessment platform for pharma Key stats: 24 AI models | 100bn+ data points | 20+ blue chip clients | 80% time reduction Core value: Build drug B/R evaluations in seconds instead of months Regulatory basis: ICH Working Group XII, Guideline E2C(R2) Capabilities:

• Dynamic (continuous) B/R monitoring vs. static periodic assessments
• Evidence mapping & cross-indication harmonization
• Regulatory gap analysis (pre-empt CHMP questions)
• Probabilistic approval prediction
• Post-marketing RWE integration
• Pediatric evidence bridging
• Biosimilar multi-indication strategy
• Re-examination support after negative opinions

Differentiators:

• Only dynamic BRA platform (competitors are static)
• Clinician-trained models (not generic AI)
• 100bn+ data points (largest BRE database)
• 80% time reduction (independently validated)

Knowledge Base Paths

Embedded References (in this skill's references/ folder)

• references/arcascience-context.md — Full company context & positioning
• references/sales-package-platform.md — BRA platform deck, one-pagers, evidence infrastructure
• references/sales-package-proposals.md — Client proposals (Dilon, Sanofi, Asphalion) & BRA analysis
• references/sales-package-personae.md — Buyer personae analysis (14 personas with scoring)
• references/sales-package-operations.md — Sales org process, contracts templates

External Dossiers (in the GitHub repo — read on demand)

• Company dossiers: C:\Users\Romai\OneDrive\Documents\GitHub\Decks-Docs\arcascience_dossiers\companies\ (100 files)
• Stakeholder profiles: C:\Users\Romai\OneDrive\Documents\GitHub\Decks-Docs\arcascience_dossiers\stakeholders\
• Master targeting table: C:\Users\Romai\OneDrive\Documents\GitHub\Decks-Docs\arcascience_dossiers\master_targeting_table.csv
• Company list: C:\Users\Romai\OneDrive\Documents\GitHub\Decks-Docs\arcascience_dossiers\COMPANY_LIST.md

Quality Checklist

• [ ] Language matches incoming email
• [ ] Tone matches sender type
• [ ] Company dossier referenced if available
• [ ] No email sent without explicit "envoyer/send" confirmation
• [ ] Specific ArcaScience capabilities mentioned (not generic)
• [ ] Call to action included for business emails
• [ ] Relevant PJ suggested for prospect/client emails (one-pager, proposal, etc.)
• [ ] PJ mentioned in email body when attached
• [ ] No contracts/CDA sent to cold prospects
• [ ] Max 3 attachments per email