Available Context & Tools

@_platform-references/org-variables.md @_platform-references/capabilities.md

Follow-Up Reply Drafter

Goal

Draft contextual reply emails for threads needing response, with suggested subject lines and clear CTAs. Every reply should feel like it was written by the rep, not a machine — it acknowledges context, advances the conversation, and asks for exactly one thing.

Why Reply Quality Matters

The data is clear: how you reply matters as much as whether you reply at all.

Replies under 100 words get 2x the response rate compared to replies over 200 words (Boomerang, 2023 analysis of 300K+ email threads).
Emails with a single CTA have 371% higher click rates than those with multiple asks (WordStream).
Personalized subject lines increase open rates by 26% (Campaign Monitor), and reply threading (keeping "Re:") gets 93% open rates.
The first sentence determines read-through: 65% of recipients decide whether to read the full email based on the opening line (Litmus).
Tone-matched replies get 40% faster responses: When your reply mirrors the formality level of the sender, they respond faster (Gong.io communication study).
Including a specific time in your CTA (e.g., "Does Thursday at 2pm work?") increases booking rates by 3.5x compared to "Let me know when you're free" (Calendly data).

Bad follow-ups don't just fail to advance the deal — they actively damage it. A generic "just checking in" email signals low effort and makes the prospect less likely to engage in the future.

Required Capabilities

Email: To access thread history and draft replies
CRM: To enrich replies with deal context and contact data

Inputs

threads_needing_response: output from followup-triage (or a single thread object)
contact_data: from execute_action("get_contact", { id }) for each thread's contact_id
(Optional) deal_data: from execute_action("get_deal", { id }) for deal-linked threads
(Optional) tone: override tone preference ("professional", "friendly", "executive")

Data Gathering (via execute_action)

Fetch thread details: execute_action("search_emails", { thread_id }) for full thread context
Fetch contact: execute_action("get_contact", { id: contact_id }) for name, title, company
Fetch deal: execute_action("get_deal", { id: deal_id }) if thread is deal-linked
Fetch recent activities: execute_action("get_activities", { contact_id, limit: 5 }) for interaction history

Context-Aware Reply Frameworks

Consult references/reply-frameworks.md for the complete framework library with 3 example replies per framework, a framework selection decision tree, CTA options, and anti-patterns to avoid.

Every reply falls into one of these scenario types. Identify the scenario first, then apply the matching framework.

Scenario 1: Answering a Question

The prospect asked something specific and you need to respond.

Framework: Answer-Expand-Advance

Answer: Direct answer to their question in the first sentence. No preamble.
Expand: One sentence of helpful context or a resource link if relevant.
Advance: Move the conversation forward with a related question or next step.

Example:

Hi Sarah,

Yes, our API fully supports OAuth 2.0 PKCE for mobile and SPA clients. Here's our auth documentation with implementation examples: [link]

Would it be helpful to schedule a 20-minute technical walkthrough with our solutions engineer this week? I have Thursday 2-3pm open.

Best,
[Rep]

Anti-pattern (what NOT to write):

Hi Sarah,

Thanks for reaching out! Great question. I wanted to circle back on your inquiry about OAuth support...
[3 paragraphs of background nobody asked for]
Let me know if you have any other questions!

Scenario 2: Delivering a Promise

You committed to sending something and now need to follow through.

Framework: Deliver-Context-Bridge

Deliver: Lead with the deliverable. Attach it or link it immediately.
Context: One sentence explaining what they will find and why it matters to them specifically.
Bridge: Connect the deliverable to the next step in the buying process.

Example:

Hi Mike,

Attached is the enterprise pricing breakdown we discussed. I've highlighted the volume tier that matches your 500-seat deployment — it comes in 18% under your current vendor.

Shall I walk through the ROI model with your procurement team? Happy to join your internal review.

Best,
[Rep]

Scenario 3: Re-engaging a Stale Thread

The conversation went cold and you need to restart momentum without being pushy.

Framework: Value-Hook-Easy-Ask

Value: Lead with something genuinely useful — a relevant insight, resource, case study, or market update. Never lead with "just checking in." Pull from ${company_name} case studies or customer success stories in the Organization Context when a relevant example exists for the prospect's industry.
Hook: Connect the value to their specific situation or a previous conversation point.
Easy Ask: Make the CTA low-friction. Not "let's schedule a call" but "worth a look?"

Warm stale (3-7 days):

Hi Lisa,

Thought you'd find this relevant — [Company in their industry] just published their Q3 results showing a 34% improvement in pipeline velocity after implementing [relevant capability]. Reminds me of the conversion challenges you mentioned.

Worth a quick look? Happy to share how the approach maps to your setup.

Best,
[Rep]

Cold stale (8-14 days):

Hi Lisa,

I came across [specific insight] and immediately thought of your team's [specific challenge they mentioned]. We just helped [similar company] solve this — here's a 2-minute case study: [link]

No pressure at all — just thought it might be useful as you evaluate options.

Best,
[Rep]

Dead stale (15+ days) — Pattern Interrupt:

Hi Lisa,

I realize I may have been approaching this from the wrong angle. Instead of [previous topic], I'm curious: what's the single biggest pipeline challenge keeping your team up at night this quarter?

Even a one-line reply would help me understand if there's a way I can actually be useful.

Best,
[Rep]

Scenario 4: Advancing a Deal

The thread is active and you need to push toward the next milestone.

Framework: Acknowledge-Advance-Commit

Acknowledge: Reference the last interaction or decision point.
Advance: Propose the specific next step with reasoning.
Commit: Offer concrete times or actions to remove friction.

Example:

Hi James,

Great call yesterday — your team's questions about data migration were exactly the right ones to raise at this stage.

Based on what we covered, I think the logical next step is a 45-minute technical deep dive with your data engineering lead. That way we can map your current schema to our migration tooling before you commit budget.

I have Tuesday 10am or Wednesday 2pm open. Which works better for your team?

Best,
[Rep]

The "Acknowledge-Advance" Pattern

This is the universal principle underlying all reply frameworks. Every good reply does two things:

Acknowledge what they said or did — proves you read their message, validates their concern, and shows you are paying attention. This is not fluff; it is trust-building.
Advance the conversation — move it toward a specific outcome. Never leave a thread in the same state you found it.

Acknowledgment Techniques

Mirror their language: If they said "we're struggling with pipeline visibility," use "pipeline visibility" in your reply, not "sales analytics."
Reference specifics: Quote a number, a name, or a date they mentioned.
Validate their concern: "That's a fair point" or "Makes sense given your timeline."

Advancement Techniques

Propose a specific next step (not "let me know your thoughts").
Offer two options (creates a choice, not a yes/no decision).
Set a deadline or date (adds productive urgency without pressure).

Tone Matching Methodology

See references/tone-matching.md for the complete tone calibration guide with sender detection signals, the tone matching matrix, the formality ladder with word substitutions, cultural and industry adjustments, and data on tone matching impact on reply rates.

Organization Context: Use the brand voice and communication style from the Organization Context to calibrate reply tone. The organization's preferred voice (e.g., authoritative, consultative, conversational) sets the baseline; tone matching with the sender adjusts from there. Reference ${company_name} value propositions and case studies from the Organization Context when they naturally support re-engagement or deal advancement.

Reply tone should mirror the sender's formality level, adjusted one notch toward professional. This builds rapport without being jarring.

Tone Detection Signals

Signal	Formal	Casual
Greeting	"Dear [Name]"	"Hey [Name]" or no greeting
Sentence structure	Complete sentences, proper punctuation	Fragments, dashes, ellipses
Sign-off	"Best regards," "Sincerely,"	"Thanks!", "Cheers," or none
Word choice	"I would like to inquire"	"Quick question"
Emoji/exclamation	None	Present

Tone Matching Rules

If they write formally: Match their formality. Use "Dear" or "Hi [Full Name]."
If they write casually: Be warm but slightly more polished. Use "Hi [First Name]" not "Hey."
If they use humor: You can acknowledge it lightly but keep it brief. Never force humor.
If uncertain: Default to "professional-warm" — friendly but not casual. "Hi [First Name]," complete sentences, "Best,"
C-suite rule: Always lean one notch more formal when writing to VP+ level, regardless of their tone.

Tone Override

If the user explicitly specifies a tone ("make it friendly," "keep it executive"), that overrides the matching algorithm. Map to:

"friendly": First name, warm opening, conversational language, exclamation mark allowed
"professional": Full name first time, measured language, structured paragraphs
"executive": Shortest possible, bullet points, no pleasantries, bottom-line-first

CTA Design for Replies

Reply CTAs are fundamentally different from cold outreach CTAs. In a reply, you already have context and relationship. Use it.

CTA Principles for Replies

Single CTA only: Never ask for two things. Pick the most important next action.
Be specific: "Does Thursday at 2pm work?" beats "Let me know when you're free."
Reduce friction: Offer to do the work. "I'll send a calendar invite" vs "Please find a time."
Match the stakes: Small ask for early-stage, bigger commitment for late-stage.
Give an out: "No pressure if the timing isn't right" — paradoxically increases response rates by 22% (HubSpot A/B tests).

CTA by Thread Category

Category	CTA Style	Example
Answering a question	Offer related help	"Want me to set up a walkthrough?"
Delivering a promise	Bridge to next step	"Shall I loop in your [role] to review?"
Re-engaging stale	Low-friction value	"Worth a look?"
Advancing a deal	Specific commitment	"Tuesday 10am or Wednesday 2pm?"
Relationship maintenance	Open-ended personal	"Would love to catch up — coffee next week?"

Thread Continuation Principles

When drafting a reply within an existing thread:

Never restart: Do not re-introduce yourself or recap the entire thread. They know who you are.
Build momentum: Reference the most recent message, not the first one.
Keep the thread: Reply in-thread, do not start a new email. Threading has 93% open rates.
Shorten over time: As the thread progresses, replies should get shorter, not longer. First reply might be 100 words. Fifth reply might be 30.
Match cadence: If they reply in 2-hour cycles, you should too. If they reply in 24-hour cycles, give them at least 12 hours.

Subject Line Modification Rules

When to Keep "Re: [Original Subject]"

Thread has < 10 messages
Original subject is still relevant
Conversation has not changed topics
You want to maintain the high open rate of threaded replies (93%)

When to Modify the Subject

Thread has changed topics significantly
Original subject is vague (e.g., "Re: Quick question" after 15 exchanges)
You are delivering a specific deliverable (change to "Re: [Original] — Pricing Attached")
Thread has > 10 messages and you want to reset readability

Subject Line Modification Format

Always preserve the "Re:" prefix for threading. Append a clarifier:

Re: Q3 Planning — Pricing Breakdown Attached
Re: Technical Review — OAuth 2.0 Details
Re: Partnership Discussion — Next Steps for Tuesday

Task Creation from Email Context

For each reply draft, generate 1-2 follow-up task previews. These ensure the rep does not just send the reply and forget about it.

Task Design Rules

One internal task: Something the rep or their team needs to do (prepare a doc, brief a colleague, update CRM).
One customer-facing task: The expected follow-up after the reply (confirm meeting, check if they opened the attachment, send the next artifact).
Due dates should be specific: "Tomorrow by 2pm" or "Wednesday EOD" — not "soon" or "next week."
Priority matches thread urgency: High-urgency thread = high-priority tasks.
Tasks must be actionable: Start with a verb. "Send," "Prepare," "Follow up," "Update," "Schedule."

Example Task Previews

Task 1 (Internal): "Update Acme deal stage to Negotiation in CRM"
Due: Today EOD | Priority: High | Deal: Acme Corp Expansion

Task 2 (Customer-facing): "Follow up with Sarah if no reply to pricing by Thursday"
Due: Thursday 10am | Priority: High | Contact: Sarah Chen

Output Contract

Return a SkillResult with:

`data.reply_drafts`

Array of 3-5 email drafts (one per top-priority thread), sorted by urgency. Each entry:

to: string (contact email)
contact_name: string
subject: string (suggested subject line, preserving "Re:" when appropriate)
body: string (the full reply text, ready to send)
body_html: string | null (HTML formatted version if available)
scenario: "answering_question" | "delivering_promise" | "re_engaging" | "advancing_deal" | "relationship_maintenance"
framework_used: string (name of the framework applied)
tone: "professional" | "friendly" | "executive"
word_count: number (must be <= 150 for standard replies)
cta: string (the specific call-to-action used)
cta_type: "specific_time" | "binary_choice" | "low_friction" | "open_ended"
thread_id: string | null
contact_id: string | null
deal_id: string | null
context_used: string (brief note on what context informed the reply)

`data.task_previews`

Array of 2-3 task previews. Each entry:

title: string (starts with a verb)
description: string (includes checklist items if multi-step)
due_date: string (ISO date, or relative like "tomorrow 2pm")
priority: "high" | "medium" | "low"
type: "internal" | "customer_facing"
contact_id: string | null
deal_id: string | null
trigger_condition: string | null (e.g., "if no reply by Thursday" — for conditional follow-up tasks)

`data.summary`

String: Human-readable summary. Example: "Drafted 4 replies: pricing follow-up to Sarah Chen (High), technical answer to Mike Rodriguez (High), re-engagement to Lisa Park (Medium), and relationship touch to James Kim (Low). Created 3 follow-up tasks."

Quality Checklist

Before returning results, validate every reply draft against these criteria:

Content Quality

Reply is under 150 words (hard limit for standard replies; executive tone allows up to 80 words)
Reply has exactly ONE call-to-action (never two)
Reply acknowledges something specific from the previous message (a quote, a number, a name, a concern)
Reply advances the conversation (thread state after sending is different from before)
Reply does not start with "Just checking in," "Hope you're well," "I wanted to follow up," or any variant of these dead phrases
Reply does not re-introduce the rep or the company (they already know)

Tone Quality

Tone matches the sender's formality level (within one notch)
No forced humor or emojis unless the sender uses them
First name used (not full name) unless writing to C-suite for the first time

CTA Quality

CTA is specific and actionable (not "let me know your thoughts")
CTA includes a specific time or option when requesting a meeting
CTA matches the thread stage (low friction for early stage, commitment for late stage)
CTA gives an easy out for re-engagement emails ("no pressure" or equivalent)

Context Quality

Deal name and stage are referenced naturally (not forced) in deal-linked replies
No confidential information from CRM is exposed to external recipients
Company name and contact title are correct and current
Dates and deadlines mentioned are realistic and not in the past
Reply tone aligns with ${company_name} brand voice from Organization Context
Case studies or value propositions from Organization Context are used where they naturally fit (especially for re-engagement and deal advancement scenarios)

Thread Quality

Subject line preserves "Re:" threading unless topic changed significantly
Reply fits naturally as the next message in the thread
Reply is shorter than or equal in length to the previous exchange

Error Handling

Missing thread context

If the thread history is unavailable, draft a general reply based on the subject line and contact data. Flag in output: "Limited thread context — reply may need manual adjustment before sending."

Contact not found in CRM

If get_contact returns null, use the email address and any name from the thread. Set contact_id: null and note: "Contact not in CRM — consider creating a record."

No deal linked

If no deal is linked, omit deal references from the reply. This is not an error — many legitimate replies are pre-deal or relationship-based.

Ambiguous scenario

If the thread fits multiple scenarios (e.g., answering a question AND delivering a promise), pick the higher-priority scenario (delivering a promise outranks answering a question). Note both in context_used.

Thread too long

If the thread has 20+ messages, focus on the last 3-5 exchanges for context. Do not attempt to summarize the entire history in the reply.

Sensitive content detected

If the thread contains pricing, contract terms, legal language, or competitor mentions, add a flag: "sensitivity": "high" and note: "Contains sensitive content — review carefully before sending."

Tone conflict

If the user requests a tone that conflicts with the thread context (e.g., "make it casual" for a C-suite negotiation thread), honor the user's request but add a note: "Requested tone (casual) differs from typical formality for this contact level (executive). Applied as requested."

Examples

Good Reply Draft

Subject: Re: Q3 Planning — Pricing Breakdown Attached

Hi Sarah,

Attached is the enterprise pricing for your 500-seat deployment — I've highlighted
the volume tier that saves 18% vs. your current setup.

Two things to flag: the annual commitment unlocks an additional 10% discount,
and we can phase the rollout across your 3 offices to reduce migration risk.

Would it help to walk through this with your procurement lead? I have Thursday
2-3pm or Friday 10-11am open.

Best,
[Rep]

Why this is good: 73 words. Delivers the promise immediately. Adds value (discount detail, phased rollout). Specific CTA with two time options. References their situation (500 seats, 3 offices).

Bad Reply Draft (what to avoid)

Subject: Re: Q3 Planning

Hi Sarah,

I hope this email finds you well! I wanted to follow up on our previous
conversation about pricing. As discussed, I'm attaching the pricing document
for your review. Our enterprise plan offers flexible pricing options for
organizations of all sizes. We have three tiers: Starter, Professional, and
Enterprise. The Enterprise tier includes unlimited users, priority support,
24/7 monitoring, advanced analytics, custom integrations, and dedicated
account management.

Please let me know if you have any questions or if you'd like to schedule
a call to discuss further. I'm happy to help in any way I can!

Looking forward to hearing from you.

Best regards,
[Rep]

Why this is bad: 118 words of fluff. Starts with "I hope this email finds you well." Generic feature list nobody asked for. Vague CTA ("let me know"). Does not reference her specific situation. Does not acknowledge what was promised.