Schedule Social Posts

Method by Federico De Ponte.

The scheduler is the last step of the content machine: a post reaches it only after it passed post-draft-gate and the owner approved the exact text. For finished vertical reels, reel-machine has its own scheduler with a pace check; this skill is for text, image, carousel and link posts.

Setup (optional, once, by the user)

No Buffer? Skip setup. schedule then sends nothing and prints the post (channel, time, text, first comment) for the user to paste by hand. Never stop or ask for a token because Buffer is missing.

With Buffer:

1. Connect the channels in Buffer (buffer.com).
2. Create an API token in Buffer's settings and export it in the user's own terminal or secret manager as BUFFER_TOKEN. Never ask for the token in chat, never put it in a command argument, a file in the repo or a log.
3. python3 scripts/buffer_posts.py channels lists the connected channels with handle and id.

Buffer's API has a daily request budget per token. Each command costs 2 to 4 requests; do not poll.

Approval rule

• schedule saves a draft in Buffer unless --armed is passed. Draft first, show the owner the text, the channel and the time, then arm it.
• A general "post the launch stuff this week" is not approval of a specific text. Arm only text the owner has seen.
• There is no publish-now command. A post always goes into a slot at least 10 minutes ahead. "Post now" means: the next open slot, shown to the owner.
• If the text changes after approval, it needs approval again.

Commands

python3 scripts/buffer_posts.py channels
python3 scripts/buffer_posts.py queue --days 14
python3 scripts/buffer_posts.py slots --channel <handle> --days 14 --cap 1
python3 scripts/buffer_posts.py schedule --channel <handle> --text-file post.txt \
    --at 2026-11-03T08:00:00-08:00 --comment "link or hashtags" --dry-run
python3 scripts/buffer_posts.py schedule ... --armed      # only after the owner approved this text
python3 scripts/buffer_posts.py delete --id <postId>
python3 scripts/buffer_posts.py analytics --days 30

• --at takes a time with its timezone offset. Use the owner's timezone; say the time in it.
• --comment goes where each platform supports it: LinkedIn, Instagram and Facebook get a real first comment; X, Threads, Bluesky and Mastodon get a reply in a thread under the post. Repeat --comment for a longer X thread.
• --image takes one public https URL, or several separated by commas for a carousel. --video takes one public https URL. Buffer has no upload: it fetches the file when the post fires, so the URL must stay reachable until then.
• Instagram posts get a type automatically (reel for video). YouTube needs --title.
• --dry-run prints the exact request and sends nothing.

Cadence

• --cap is the most posts per channel per calendar day (UTC); default 1. Drafts count as taken, because a draft is a reservation the owner may arm later.
• schedule refuses a day that is full and points to slots. Never raise the cap on the owner's behalf.
• The check reads Buffer's queue only. If the owner also schedules on the same account with another tool, those posts are invisible here: ask, and count them by hand.
• Before moving or replacing any post, read its text for dates and words like today, tomorrow, this week or an event name. A date-bound post stays on its day.

After posting

analytics reports Buffer's aggregated metrics per channel. Add the per-post numbers to the post-signals sheet so the next idea comes from what actually worked.