Two posts covering nearly the same thing is a common mess: they split your search ranking, and a reader who lands on one never sees the other. The fix is to fold them into a single stronger post and retire the loser, but nobody wants to hand-stitch two bodies together and then second-guess which URL to keep.
This reads both, proposes one merged version, and only trashes the leftover after you have signed off.
I have two posts that overlap and I want to merge them: [/post-a/] and[/post-b/]. Do not change anything yet.
Read both in full. Then propose a single merged post: combine the two bodiesinto one, drop the repetition, keep the best of each, and tell me which postyou would keep as the home for the merged content and which you would trash.Show me the merged copy and your reasoning. Change nothing until I approveboth the copy and which post to keep.
Once I approve, update the keeper with the merged body, then move the otherpost to the Trash. Do not permanently delete anything. When you are done, tellme what you changed, confirm the trashed post can be restored, and remind methe keeper's old version is saved as a revision.Why this is safe to run
Reading both posts writes nothing, so you see the full merged draft and decide which URL survives before anything is saved or removed.
The two writes are deliberately split. Updating the keeper only edits its body and never changes its published status, and its previous version is saved as a revision. Trashing the other post moves it to the Trash rather than deleting it, so it stays recoverable until you empty the Trash yourself. Nothing here bypasses that safety net. Each ability is off until you enable it, every call is checked against the capability of the user you bound the agent to, and both the update and the trash land in the Activity Log with the post ids, so you can see exactly what happened.
Already connected? Skip the paste. Ask your agent for the “merge two posts into one” recipe, name the two posts, and it runs the same read-propose-approve flow on your own site.