Run the phase1 planner suites' tool steps in auto mode - #14
Closed
Parad0x-Labs wants to merge 2 commits into
Closed
Parad0x-Labs wants to merge 2 commits into
Parad0x-Labs wants to merge 2 commits into
Conversation
added 2 commits
September 21, 2026 09:51
The execution gate's tightened hands-off law (only proven read-only commands run unprompted outside Auto) means a modeless context is MANUAL, so every workspace.run_tests / apply step in these suites refused with user_action_required before the planner ever saw a failed validation -- 16 operator-phase1 and 8 orchestration-phase1 failures locally and in CI, with the planner honestly falling back to workspace_stop_after_read. The green sibling (test_code_task_reviewed_base) already passes operating_mode auto; these suites now do the same at every envelope and source_context site. The approval law itself stays policed by the permission-authority suites.
… gated write The two manual-mode sabotage tests set manual BEFORE the journey, so the tightened hands-off law refused the repro step itself (user_action_required) and the drive never reached the assertion under test -- that manual mode blocks the WRITE. The journey now runs in auto and the mode switches at the mutate step, which is the behavior both tests actually pin.
This was referenced Sep 21, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Root cause
24 CI+local failures (16
test_runtime_execution_operator_phase1, 8test_orchestration_execution_phase1) all reduce to one shape: the suites drive the planner state machine with REAL tool executions, but their source contexts carry no operating mode. Under the execution gate's tightened hands-off law (core/execution_gate.py: only proven read-only commands run unprompted outside Auto), a modeless context is MANUAL — everyworkspace.run_tests/ patch step refused withuser_action_requiredbefore the planner ever saw a failed validation, and the planner honestly fell back toworkspace_stop_after_readwhere the tests expectedplanned_*continuations.Repair
The suites now pass
operating_mode: autoat every source-context and envelope-execution site — the same contract the green sibling (test_code_task_reviewed_base.py:74) already uses. The approval law itself is untouched and stays policed by the permission-authority suites; these tests exercise planning behavior, not approval behavior.Evidence
tests/test_runtime_execution_operator_phase1.py: 16 failed → 25 passed;tests/test_orchestration_execution_phase1.py: 8 failed → 17 passed; both together 42 passed; ruff clean.status=user_action_required→_latest_failed_validation_observation→ None → stop_after_read).Contributed by sls_0x.