fix(runtime): auto-continue a turn the model stalled without a tool call

A reply carrying no tool_use ended the turn unconditionally. Small local
models routinely narrate the next tool call in prose — "Now running the
EEC-V probe:" — and then emit EOS, so the announced work never happened
and the user was dropped back at the prompt. The same failure wears a
second hat mid-task: stopping to ask for a go-ahead on work already
requested.

The loop now inspects the final line of a tool-less reply. If it
announces a pending action, or asks a question after tools have already
run this turn, a short nudge is pushed as a user message and the loop
continues instead of breaking. The budget is bounded (default 3 per
turn, CLAW_MAX_STALL_NUDGES, 0 disables) so a model that really is
finished still gets to stop.

Prevention pairs with recovery: the "Doing tasks" system-prompt section
now tells the model to act in the turn it plans in.

Tests cover the auto-continue, the bound, a plain completion that must
NOT be nudged, the disable switch, and last-line-only detection. The
behavioural pair was mutation-checked: reverting the continue to a break
fails both.

(cherry picked from commit 9035fe6f4bf4a8d4cb70d7a3baf1f3748c757498)
This commit is contained in:
reacechalmers authored and clawdbot committed 2026-10-12 22:58:32 +00:00
1 parent cb572811e1
commit 7c38d78759
2 files changed
+251 -3

No files matched your search

+1
View File
@@ -783,6 +783,7 @@ fn get_simple_doing_tasks_section() -> String {
"Do not add speculative abstractions, compatibility shims, or unrelated cleanup.".to_string(),
"Do not create files unless they are required to complete the task.".to_string(),
"If an approach fails, diagnose the failure before switching tactics.".to_string(),
"Act in the turn you plan in: when the next step is a tool call, make the call. Never end a turn describing a step you have not taken, and do not stop to ask for a go-ahead on work already requested.".to_string(),
"Be careful not to introduce security vulnerabilities such as command injection, XSS, or SQL injection.".to_string(),
"Report outcomes faithfully: if verification fails or was not run, say so explicitly.".to_string(),
]);