ihuzaifashoukat / x-use-content

x-use content

Research -> draft -> polish -> stage. Ordinary post tools stage only when
mcp.draft_mode is enabled (default); with it disabled they publish immediately.

Workflow

1. Load the voice. get_account(account), read persona and
   target_keywords. No persona set? Ask the user a couple of questions and
   set one via update_account(account, persona=...) first (see x-use-setup).

2. Research the niche. search_tweets per keyword (limit 5). Note
   which formats earn engagement: hot takes, how-tos, charts, contrarian
   questions, build-in-public numbers. get_tweet on standouts to see their
   media.

3. Pick 3 angles. Tell the user the angles you see and let them steer.

4. Draft one at a time. Write the post yourself in the persona (at most
   280 chars; do not assume long-form account support), then
   post_tweet(account, text=...) to stage it as a draft. Show it. Take the
   user's edits and re-stage if needed; reject the superseded pending draft to
   avoid approving the old text later. Polish each draft before the next.

5. Review. End with list_drafts(account) and let the user choose:
   approve_draft to post now, queue_post-style scheduling via
   queue_post(account=..., text=..., not_before=<ISO 8601 timestamp>) for later,
   or reject_draft. Scheduling stores a separate final payload; inspect it
   via list_queue and reject the superseded draft. Due work runs through
   process_queue, or the opt-in auto-drain worker; enqueueing is not approval
   to publish unless the user's scheduling authorization covers it.

Rules

• Original text in the user's voice. Never repost someone else's wording.

• Media for posts: only local file paths the user provides
  (post_tweet(..., media=[path])). Never invent image files.

• For X thread composition, use x-use-threads so the sequence is staged and
  its durable progress can be reviewed safely.

• Server-side generation (generate_and_post, draft_post_variations)
  needs the configured llm block. Prefer explicit text for interactive work;
  no server LLM key is needed when you write it yourself.