Repository navigation
Conversation
…ODELTRACE_RETRY_COUNT environment variables
…ry from enabled flag
…n reaching interval
…AUTO_ACTIVATE for auto probe
…eshold=2) When SQLite telemetry detects server-side model downgrade (Safety Buffering) on consecutive checks, the plugin now triggers a hard taskHalt that the AI cannot self-clear — only the user can resume via 'guard.mjs resume --halt <id>'. Changes: - Default retryCount 3→2 (fingerprint probe threshold) - Add consecutiveDegradedCount to session state for tracking - controlContext() handles telemetry-sourced halts with specific messaging - handleHook() SQLite block: count consecutive degradations, trigger taskHalt at threshold (default 2, configurable via MODELTRACE_TELEMETRY_HALT_COUNT) - queueTelemetryAlert() accepts consecutiveCount/halted options for severe alerts - resume command resets consecutiveDegradedCount - New test: two consecutive telemetry degradations trigger taskHalt hard stop - Existing tests updated for retryCount=2 default
… for telemetry degradation
…r waiting in degraded session
- Eliminate continuation dead loop between Stop hook (decision: block) and Codex autonomous active goal scheduling - Deliver halt instructions once and avoid returning block on subsequent turns when task is halted - Pause host thread goal via native thread/goal/set JSON-RPC interface and verify state transitions atomically - Enforce strict restricted control channel during task halt, allowing status inspection while blocking task work - Require explicit halt ID on resume and verify host goal unpausing before clearing halt state - Deduplicate telemetry alerts during halt and prevent healthy telemetry from auto-clearing non-canary halts - Add multi-turn regression test suite covering loop prevention, status transitions, lock contention, and stdio hook execution
… track subagents - Extend SQLite telemetry checking to PreToolUse and PostToolUse hooks for real-time circuit breaking in continuous turns - Persist subagentSessionIds to enable bidirectional cross-thread telemetry correlation - Resolve parent session in busyHookFallback under lock contention to prevent subagents from bypassing task halt - Guard alert turn delivery to prevent internal subagent hooks from burning user alert turns - Enforce strict session isolation in queryLatestRecord when allowGlobalFallback is false - Add regression tests covering continuous-turn halt, subagent lock contention, cross-thread detection, and alert delivery
imbajin
changed the title
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
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.
Closed by mistake.