tweet-broadcast

name	tweet-broadcast
description	Post queued tweets and replies from .pending-x/ via X API v2 (tweepy)
var
tags	["twitter","delivery"]

PAUSE CHECK — run this first

Read memory/goals.json. If tweetingPaused is true, log "tweet-broadcast: paused by operator — skipping all posts" to memory/logs/{today}.md and exit immediately. Do not post anything, do not move files.

---

Post queued tweet files from .pending-x/ to Twitter/X using the X API v2.

CRITICAL: Never thread tweets

Each tweet-*.txt file is a standalone post. Never pass --reply-to for tweet files. Do NOT chain tweets into threads. Each one goes out independently to the main feed. Only reply-{TWEET_ID}.txt files use --reply-to. Threads appear under "Replies" not "Posts" — this is wrong.

Check for queued content

ls .pending-x/*.txt 2>/dev/null | head -20

If no files exist, log "tweet-broadcast: nothing queued" and exit cleanly.

Dedup gate — before any X API call

For each queued file, check whether it was already posted by looking up its basename in memory/x-tweet-log.jsonl before calling X:

import json, os
posted = set()
try:
    with open("memory/x-tweet-log.jsonl") as f:
        posted = {json.loads(l)["source_file"] for l in f if l.strip()}
except FileNotFoundError:
    pass
# skip if os.path.basename(filepath) in posted

If already posted: move the file to .pending-x/sent/ (use the Write tool if mv fails in sandbox) and skip — no X API call. This prevents duplicate-content 403s and saves POST quota.

Rate ceiling

Process at most 5 files per run. Tweets (tweet-*.txt) before replies (reply-*.txt). Sort by filename (chronological order).

Post each file

For each file (up to 5):

1. Strip the #content_type: first line — that is metadata only, never tweet text.
2. For tweet files (tweet-*.txt) — standalone post, no reply-to:

   python scripts/tweet-browser.py --action post --file .pending-x/tweet-20260604-on-chain-report.txt
   

3. For reply files (reply-{TWEET_ID}.txt) — reply to that tweet ID:

   python scripts/tweet-browser.py --action post --file .pending-x/reply-1234567890.txt --reply-to 1234567890
   

4. Parse stdout JSON. On "status": "ok":
   • Move file to .pending-x/sent/
   • Append to memory/x-tweet-log.jsonl:

     {"tweet_id":"1234567890123456789","content_type":"on-chain-report","text":"0.485 DIEM/day...","posted_at":"2026-06-04T14:00:00Z","source_file":"tweet-20260604-140000-on-chain-report.txt"}
     

5. On "status": "error": leave file in .pending-x/, log error, continue.

After posting

Log to memory/logs/{today}.md:

tweet-broadcast: posted N tweet(s), N replies — N failed (left in queue)