[FEATURE]: Make doom_loop repeat threshold configurable

[jjangga0214]

Feature hasn't been suggested before

• I have verified this feature I'm about to request hasn't been suggested before.

Describe the enhancement you want to request

permission.doom_loop is useful, but the current behavior appears to be hard-coded to trigger when the same tool call repeats 3 times with identical input. It would be helpful to make that repeat threshold configurable in opencode.json instead of fixed.

A configurable threshold would let users tune the guard for different workflows:

• lower values for tighter cost/safety control in autonomous runs
• higher values for iterative/debugging workflows where a few repeated retries are expected

A good direction would be to let permission.doom_loop accept staged thresholds directly:

{
  "$schema": "https://opencode.ai/config.json",
  "permission": {
    "doom_loop": {
      "4": "ask",
      "6": "deny"
    }
  }
}

That would allow progressive escalation instead of one fixed cutoff:

• 4+ repeated identical tool calls => ask
• 6+ repeated identical tool calls => deny

This feels more flexible than adding a separate top-level threshold field, because it keeps the behavior inside the existing permission system and supports multi-stage escalation naturally.

The exact config shape is up to you, but exposing the threshold this way would make the existing doom-loop protection much more practical.

[Keesan12]

Configurable thresholds would help, but I’d avoid making one magic number do all the work.

The same repeat count can mean very different things depending on whether anything changed. 4 identical calls + progress is usually fine; 4 identical calls + no file/state/verifier movement is the real danger. If you add staged escalation, I’d include that progress bit in the decision instead of only exposing a bigger/smaller threshold knob.

[github-actions]

To stay organized issues are automatically closed after 60 days of no activity. If the issue is still relevant please open a new one.