model = "gpt-5.3-codex-spark"
model_reasoning_effort = "medium"
personality = "pragmatic"
tool_output_token_limit = 25000
# Leave room for native compaction near the 272–273k context window.
# Formula: 273000 - (tool_output_token_limit + 15000)
# With tool_output_token_limit=25000 ⇒ 273000 - (25000 + 15000) = 233000
model_auto_compact_token_limit = 233000
[features]
ghost_commit = false
unified_exec = true
apply_patch_freeform = true
web_search_request = true
skills = true
shell_snapshot = true