# Bl Thread

> Handle the comment thread under the user's own LinkedIn post - sort every comment by what it is worth, then draft the replies in that order. Use when the user pastes the comments on their post, says "reply to these", "handle my comments", "someone said X on my post", or is dealing with a critic or a lead in their own thread.

- Skill: `ehteshambuildagents/bl-thread` (Agent Skill)
- Install (CLI): `npx skillmds@latest add ehteshambuildagents/bl-thread`
- Raw SKILL.md: https://api.skillmd.com/api/skills/ehteshambuildagents/bl-thread/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: ehteshambuildagents (https://skillmd.com/u/ehteshambuildagents)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/ehteshambuildagents/bl-thread

---


# bl-thread

The thread under a post is where its reach is decided. Every reply is a second
engagement event, and the first hour does most of the work. But the comments are
not worth the same amount, so this skill sorts before it writes.

## Input

The user pastes the comments, ideally with names and roles. Screenshots are
fine. Do not scrape the thread with a browser tool.

## Sort first, and say the counts out loud

| bucket | what it looks like | what it gets |
| --- | --- | --- |
| **LEAD** | describes a problem the user solves, or asks about working together | a real answer today, then a move to DM |
| **SUBSTANCE** | disagrees, adds evidence, or asks a hard question | the longest reply in the thread |
| **PEER** | someone worth knowing, adding their own angle | a reply that gives them something back |
| **SUPPORT** | agreement, a short story, genuine warmth | two lines, by name, no boilerplate |
| **NOISE** | "Great post", tagging, pitching, bots | a reaction, or nothing |

Tell the user the counts before writing: "14 comments: 1 lead, 3 substance, 4
peer, 5 support, 1 noise." That alone tells them how the post did, and it is
information the reaction count does not carry.

## Then write, in that order

**LEAD.** Answer the actual question first, in public, properly. Do not open
with "DM me" - answering in the thread is what makes the next person ask. Move
to DM only after the answer, and only if there is a real reason.

**SUBSTANCE.** These deserve the most effort, especially the disagreements. A
well-handled disagreement outperforms agreement, because everyone reading is
watching how the user behaves under pressure. Concede the true part first. Then
the reason you still hold the position. Never the last word for its own sake.

**PEER.** Give them something: a reference, a counter-case, a question they
would enjoy. The goal is that they remember the exchange next week.

**SUPPORT.** Use their name. Add one specific thing. "Thanks Sarah" is worse
than no reply because it is visibly automatic.

**NOISE.** A reaction. Pitches get nothing. Never argue with a bot.

## Rules

1. **Never the same opening twice in one thread.** It is all visible on one
   screen and repetition is the most obvious tell there is.
2. **Vary length.** Two words to two sentences. A thread of identical replies
   looks generated because it is.
3. **No new claims.** Every fact in a reply comes from `ledger.md`, same as a
   post.
4. **Answer the question that was asked**, not the one that is easier.
5. **Hostile comments**: answer the strongest version of their point, once, then
   stop. Do not reply twice to someone looking for a fight.

## Before showing them

```bash
cd ../bl-scrub && python3 scrub.py replies.txt --explain
```

Then deliver each reply in its own block, labelled with the comment it answers
and its bucket. The user posts them.