# Twitter Social Selling

> Engage on Twitter/X for social selling — monitor keywords, find relevant threads, draft thoughtful replies, track engagement with ICP accounts. Use on "warm up prospects on Twitter", "monitor X for <keyword>", or "draft replies to these posts".

- Skill: `composio-community/twitter-social-selling` (Agent Skill)
- Install (CLI): `npx skillmds@latest add composio-community/twitter-social-selling`
- Raw SKILL.md: https://api.skillmd.com/api/skills/composio-community/twitter-social-selling/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: composio-community (https://skillmd.com/u/composio-community)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/composio-community/twitter-social-selling

---


# Twitter / X Social Selling

Warm up prospects through genuine Twitter engagement via Composio.

## Workflow

1. **Define the beat.** Ask for: keywords/hashtags to monitor, 10-20 ICP accounts to follow closely, the POV the user wants to represent.
2. **Link Twitter:**
   ```bash
   composio link twitter
   ```
3. **Monitor signals:**
   ```bash
   composio execute TWITTER_RECENT_SEARCH -d '{ "query": "\"<pain phrase>\" -is:retweet lang:en", "max_results": 25 }'
   ```
4. **For each post**, score the reply opportunity:
   - Is the author in the ICP?
   - Does the post invite dialogue (question, unclear opinion, call for advice)?
   - Can we add a genuinely useful data point, not a pitch?
5. **Draft replies** — never a product pitch. Rules:
   - < 280 chars
   - Add a specific experience, number, or counter-example
   - No CTA in the first reply; earn the DM
6. **Surface top 5 replies** for user approval, then post:
   ```bash
   composio execute TWITTER_CREATION_OF_A_POST -d '{
     "text": "<reply>",
     "reply": { "in_reply_to_tweet_id": "..." }
   }'
   ```
7. **Log engagement** (tweet_id, author, reply_text, date) to `./social-selling.csv`. Track whose replies get liked/replied to — those are your warmest prospects.

## Reply framework

- **Agree + extend:** "Yes, and we saw the same — specifically <data>."
- **Nuance:** "True above <size/stage>; below that we found <opposite>."
- **Question:** "Curious — did you find <specific subvariable> mattered?"

## Guardrails

- Max 5 replies/day — more reads as spam.
- Never reply to the same account more than once per 48h without organic signal.
- No automation of likes/follows — engagement only.
- Hard no on sending DMs before at least 2 public interactions.