Skip to content

Agent Zero AI framework

Clarify code execution status handling

Alessandrod3bec80611a8
diff --git a/plugins/_code_execution/prompts/agent.system.tool.code_exe.md b/plugins/_code_execution/prompts/agent.system.tool.code_exe.md
index e1ddf64bd..56a6f355c 100644
--- a/plugins/_code_execution/prompts/agent.system.tool.code_exe.md
+++ b/plugins/_code_execution/prompts/agent.system.tool.code_exe.md
@@ -14,7 +14,7 @@ rules:
- replace placeholder or demo data with real values before execution
- use `print()` or `console.log()` when you need explicit output
- do not interleave other tools while waiting
-- ignore framework `[SYSTEM: ...]` info in output
+- treat trailing framework `[SYSTEM: ...]` info as execution status, not command output; use it to decide whether to wait, reset, rerun, or continue
- probe cwd files tools and dependencies before expensive commands
- split long work into small commands: inspect, prepare, run, verify
- for builds installs servers training and long tests, redirect logs and poll with `runtime=output`