mostafa-drz/respond-to-message

Respond to Message

Craft responses to messages in your voice, matched to the platform. Copies to clipboard for pasting.

Preferences

Before starting, use the Read tool to read ~/.claude/skills/respond-to-message/preferences.md. If the file does not exist, treat as "no preferences set" and use defaults.

Tone Profile

Use the Read tool to read ~/.claude/skills/respond-to-message/reference/tone-profile.md. If it does not exist, fall back to preferences tone keywords. If neither exists, ask the user to describe their tone.

Platform Context

Use Glob to check which platform files exist at ~/.claude/skills/respond-to-message/reference/platforms/*.md. For the active platform, read its context file. If no file exists for the requested platform, use sensible defaults for that platform's conventions.

Command routing

Check $ARGUMENTS:

• help → display help then stop
• config → interactive setup then stop
• reset → delete ~/.claude/skills/respond-to-message/preferences.md, confirm, stop
• tone → open tone profile editing flow then stop
• anything else (including empty) → run the skill

Help

Respond to Message — Craft replies in your voice, matched to the platform

Usage:
  /respond-to-message <platform>           Craft a response (then paste the message)
  /respond-to-message                      Auto-detect platform or ask
  /respond-to-message --refine             Re-draft the last response with feedback
  /respond-to-message --formal             Override tone to more formal
  /respond-to-message --casual             Override tone to more casual
  /respond-to-message config               Set up tone, platforms, and context
  /respond-to-message tone                 Edit your tone profile with examples
  /respond-to-message reset                Clear preferences
  /respond-to-message help                 This help

Platforms: linkedin, slack, gmail, teams, whatsapp, twitter, generic

Examples:
  /respond-to-message linkedin             Then paste the LinkedIn message in chat
  /respond-to-message slack                Then paste the Slack message + optional draft
  /respond-to-message gmail --formal       Reply to an email with formal tone override

Current preferences:
  (from preferences.md or defaults)

Config

Use AskUserQuestion to collect:

Q1: Tone keywords — How would you describe your communication style? (e.g., "direct, warm, professional but not stiff, uses analogies")

Q2: Your role/title — Used for professional context in responses (e.g., "VP of Engineering at Acme Corp")

Q3: Active platforms — Which platforms do you want configured? (multiSelect: LinkedIn, Slack, Gmail, Teams, WhatsApp, Twitter/X)

Q4: Sample messages — "Would you like to paste 2-3 example messages you've written so I can learn your tone? (recommended)" (Yes/No)

If Q4 is Yes, use AskUserQuestion to collect sample messages, then analyze them for:

• Sentence length patterns
• Greeting/closing conventions
• Level of formality
• Use of humor, emojis, or directness
• Vocabulary tendencies

Save tone analysis + samples to reference/tone-profile.md.

For each platform selected in Q3, ask:

Platform-specific Q — "Any specific context for {platform}? (e.g., your LinkedIn headline, Slack workspace role, email signature)" — save to reference/platforms/{platform}.md.

Save overall preferences to ~/.claude/skills/respond-to-message/preferences.md.

Tone

Interactive flow to update the tone profile:

1. Read current reference/tone-profile.md (if exists, show summary)
2. Ask: "Paste 2-3 messages you've recently sent that represent your voice."
3. Analyze patterns and update reference/tone-profile.md
4. Confirm: "Tone profile updated. Key traits: {traits}."

Reset

Delete ~/.claude/skills/respond-to-message/preferences.md and confirm: "Preferences cleared. Using defaults. (Tone profile and platform contexts are preserved — delete them manually if needed.)"

First-time detection

If no preferences file AND no tone profile exist, show:

First time using /respond-to-message? Run /respond-to-message config to set up your tone and platforms, or just continue — I'll ask what I need as we go.

Then proceed normally.

Workflow

Step 1: Detect platform

If $ARGUMENTS contains a platform name (linkedin, slack, gmail, teams, whatsapp, twitter, generic), use it.

If no platform specified, check the pasted message for cues:

• "via LinkedIn" / LinkedIn formatting → linkedin
• Slack-style formatting (channel names, @mentions, threads) → slack
• Email headers (From:, Subject:, etc.) → gmail
• Teams-style formatting → teams

If still ambiguous, ask:

What platform is this message from?

Options: LinkedIn, Slack, Gmail, Teams, WhatsApp, Twitter/X, Other

Step 2: Parse the input

From the conversation context, extract:

• Incoming message — the message the user received and needs to respond to
• Draft response (optional) — the user's initial attempt, to be refined
• Sender context — who sent it (name, title, relationship if apparent)
• Thread context — is this a reply in an ongoing conversation?
• Specific instructions — any guidance like "decline politely" or "express interest but ask about timeline"

If the incoming message isn't clear, ask:

Paste the message you want to respond to. You can also include a draft response if you have one.

Step 3: Load tone and context

1. Tone profile — read from reference/tone-profile.md
2. Platform context — read from reference/platforms/{platform}.md
3. Preferences — read from preferences.md

Merge into a response framework:

• Voice: tone keywords + sample patterns
• Format: platform-specific conventions (see Step 5)
• Context: role, company, expertise areas

Step 4: Analyze and strategize

Before drafting, consider:

• Intent: What does the sender want? What should the response achieve?
• Relationship: Professional contact, colleague, recruiter, client, friend?
• Stakes: Casual conversation, business opportunity, sensitive topic?
• Action needed: Accept, decline, defer, negotiate, inform, ask?

If the user provided a draft, identify what to keep and what to improve.

If the intent is unclear, ask one clarifying question:

This looks like [interpretation]. Do you want to [suggested action], or something else?

Step 5: Craft the response

Write the response following platform conventions:

LinkedIn:

• Professional but personable
• Keep under 300 words for DMs
• No markdown (LinkedIn strips it) — use line breaks for structure
• Opening: acknowledge their message specifically
• Close: clear next step or warm sign-off

Slack:

• Concise, conversational
• Use emoji sparingly (match workspace culture)
• Bold for emphasis, code blocks for technical content
• Thread-aware — reference what was said above
• No formal sign-offs

Gmail:

• Proper email structure (greeting, body, sign-off)
• Match formality to the sender's style
• Keep paragraphs short (2-3 sentences)
• Clear subject line suggestion if starting a new thread
• Include email signature if configured

Teams:

• Similar to Slack but slightly more formal
• Support for mentions (@Name)
• Keep it scannable

WhatsApp:

• Short, casual
• Emoji-friendly
• Break into multiple short messages if content is long

Twitter/X:

• Under 280 characters (or note if a thread is needed)
• Punchy, direct
• Hashtags only if relevant

Generic:

• Match the sender's format and length
• Default to professional-casual

Apply tone profile throughout. The response should sound like the user, not like AI.

Step 6: Copy to clipboard

1. Write the response to /tmp/respond-to-message-output.md using the Write tool
2. Run pbcopy < /tmp/respond-to-message-output.md via Bash
3. Report:

Copied to clipboard. Ready to paste.

---

{the response, displayed for review}

/respond-to-message --refine to adjust, or just paste it.

Step 7: Handle refinement

If --refine flag is set or the user asks to adjust:

1. Ask what to change (shorter, more formal, different angle, etc.)
2. Re-draft
3. Copy new version to clipboard
4. Show diff summary: "Made it [shorter/more formal/etc.]"

Principles

• Sound like the user, not like AI — the response must match the user's actual voice. Generic, overly polished AI-speak is a failure. Use the tone profile religiously.
• Platform-native formatting — a LinkedIn DM should look like a LinkedIn DM, not a formatted email. Match conventions exactly.
• Clipboard-first — always pbcopy. The user should be able to Cmd+V immediately after the skill runs.
• Clarify when unsure — if the intent or desired action isn't clear, ask ONE focused question rather than guessing wrong.
• Minimal not maximal — shorter responses are usually better. Don't pad with pleasantries or filler unless that's the user's actual style.