Limits per Run
Every Scriptbox run has a stated, enforced budget. A limit you can see beats an outage you can't: a runaway loop dies at the deadline with its logs intact, instead of the platform killing the invocation with no record at all.
Console footer showing runtime, memory and API limits
## The per-run budget
These apply to every server-side run unless a trigger below overrides them.
| Limit |
Value |
Why |
| Wall clock (console) |
15 seconds |
Forge kills a console invocation at 25s; the gap leaves room to return logs and record the run |
| Memory |
128 MB |
QuickJS heap cap, enforced by the VM |
| Stack |
2 MB |
VM stack cap |
| REST calls |
200 |
Stops a runaway loop from hammering your site |
| Log entries |
500, each up to 10,000 characters |
Keeps responses inside Forge's payload limit |
| Result size |
512 KB |
Oversized results are replaced with a notice — never silently truncated |
25 seconds is a platform ceiling, not a preference. Past it Forge kills the invocation outright: no logs, no run history, no recorded firing.
Budgets by trigger
| Trigger |
Time budget |
Notes |
| Console run |
15s |
Also used by Run now on a scheduled job |
| Event listener |
20s per event delivery |
Shared by every listener bound to that event — three listeners get one clock between them, not three |
| Scheduled job |
60s per job |
All jobs due on one tick share a 4-minute tick budget; jobs past it are deferred to the next tick, not dropped |
| Scripted field |
10s and 50 REST calls per batch |
Jira asks for a batch of issues at once — one JQL search for the batch, never one call per issue |
| Script validator |
5s |
Someone is waiting on the transition. Fails closed: an error or timeout blocks it |
| Post function |
20s |
Runs after the transition, asynchronously |
| Jira Automation action |
45s |
Forge allows this class of invocation 55s |
| Behaviour |
1 second |
Runs in the browser while someone uses the form; no network, no timers |
A script that genuinely needs longer than the console allows belongs on a scheduled job, not in a bigger number here.
Size limits
| Item |
Limit |
| Script source |
100,000 characters |
| Script, listener, job and behaviour names |
120 characters |
One storage value |
200 KB per key — split larger data across keys |
| Scheduled-job input |
10,000 characters of JSON |
| Listener project scope |
50 projects (leave empty for every project) |
| Behaviour script |
30,000 characters; 50 affected fields; 200 project × issue type × view contexts |
| Scripted-field cache lifetime |
0 to 86,400 seconds |
| Migration analysis |
100 scripts per analysis |
Other hard edges
- Five minutes is the scheduling floor. Forge wakes the app every five minutes, so a job written as
*/1 * * * * runs once per tick.
- Types are erased, not checked. A type error is a runtime error.
- No
import. There is no module system; import is rejected with an explanation.
- No multipart. Attachment uploads and binary content cannot cross the sandbox boundary.
When a run hits a limit, the error names it. Troubleshooting lists the common ones and what to change.