Add a boolean filter_exclude parameter to the bash_output tool that inverts the filter behavior:

• When true, lines matching filter are excluded instead of kept
• Key behavior: excluded lines do NOT cause early return from timeout
• Waiting continues until non-excluded output arrives or process exits

Problem

When polling bash_output for long-running processes like wait_pr_checks.sh, agents generate excessive tool calls because:

1. Scripts emit periodic progress output (e.g., ⏳ Checks in progress... every 5s)
2. timeout_secs only sets upper bound - returns early when ANY output arrives
3. Agent sees output, polls again, sees more progress, polls again...

Solution

With filter_exclude:

bash_output(process_id="...", filter="⏳", filter_exclude=true, timeout_secs=60)

The agent can set a long timeout and only wake when meaningful output (✅ or ❌) arrives.

Tests Added

• Basic exclusion works
• Error when filter_exclude without filter
• Keeps waiting when only excluded lines arrive
• Returns when process exits (even if only excluded output)
• Timeout returns (doesn't hang) with only excluded output

Generated with mux