Post body edits are the scariest kind of content automation.
So when I automate body transforms, I try to make them safe to run twice: - run once → it changes what you expect - run again → it changes nothing
That second run is a surprisingly strong verification signal.
If you’re about to run body edits in bulk, I recommend doing this “run again” check every time.
Why body edits are risky
- Rich editors have structure (not just HTML).
- One bad “search and replace” can break formatting.
- A small bug can repeat across hundreds of posts.
The safety loop I use
1) Dry-run: show a plan (no writes). 2) Apply: write only with explicit flags. 3) Verify: re-fetch the post. 4) “Run again” check: dry-run again and confirm it would do 0 changes.
Example: “run twice” verification (what it should look like)
First run (plan):
{
"mode": "plan",
"would_change": 3,
"operations": [
{ "action": "replace", "pattern": "<old>", "to": "<new>" }
]
}
After apply, second run (plan again):
{
"mode": "plan",
"would_change": 0,
"note": "Second run: no changes needed."
}
Backups on apply
When the tool applies a body change, it should save: - a before snapshot - an after snapshot
That way, if something looks wrong later, you have a path to recover.
Get access
- Read the safety model.
- See the Ghost API tool.
