adjust tool prompt

This commit is contained in:
Concedo 2026-09-23 02:35:51 +08:00
parent 0b51371bcf
commit f2e883aeea

View file

@ -1197,7 +1197,7 @@ def review_tool_call(
"content": (
f"Tool danger classification check: Please review the proposed '{name}' function call requested immediately above, as well as the earlier request. Is it risky to execute? "
"Reply with a single word 'APPROVED' only if its full effect is clearly completely safe and harmless. "
"Otherwise, reply with a single word 'CAUTION' for any incorrect, potentially risky external side effects, credential access or disclosure, or any uncertainty. "
"Otherwise, reply with a single word 'CAUTION' for any incorrect, potentially risky external side effects, credential access or disclosure, or any task uncertainty. "
"Finally, reply with a single word 'DANGER' for any dangerous, irreversible, potentially destructive or hard to reverse actions. Remember, only reply with a single word of text!"
),
},