Upstream PR triage batch 2: 12 PRs with tests, fixes, and enhancements#368
Open
whitmo wants to merge 46 commits intodlorenc:mainfrom
Open
Upstream PR triage batch 2: 12 PRs with tests, fixes, and enhancements#368whitmo wants to merge 46 commits intodlorenc:mainfrom
whitmo wants to merge 46 commits intodlorenc:mainfrom
Conversation
Fixes branch naming inconsistency by standardizing on 'multiclaude/' prefix. Maintains backward compatibility for cleanup of legacy 'work/' branches. Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
The comment at internal/cli/cli.go:5128 referenced docs/EXTENSION_DOCUMENTATION_SUMMARY.md which doesn't exist. Updated the comment to reference the actual existing extension docs (docs/extending/SOCKET_API.md and docs/extending/STATE_FILE_INTEGRATION.md) that need to be kept in sync when CLI commands affecting extension surfaces change. Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
…space Improve the repair command to be more comprehensive by ensuring core agents and a default workspace exist after cleanup. This enhancement aligns with ROADMAP.md P1 "Agent restart" by making repair more robust and reducing the need for manual intervention. Changes: - CLI: Add ensureCoreAgents() and ensureDefaultWorkspace() helpers - CLI: Update localRepair() to recreate missing core agents - CLI: Create default workspace "my-default-2" if none exist - Daemon: Add ensureCoreAgents() and ensureDefaultWorkspace() methods - Daemon: Update handleRepairState() to recreate missing agents - Both: Improve output to show what was removed and what was created - Tests: Add comprehensive tests for all scenarios Key Features: 1. Recreates missing supervisor agent if absent 2. Recreates missing merge-queue (non-fork) or pr-shepherd (fork) 3. Creates default workspace if no workspaces exist 4. Does not duplicate existing agents/workspaces 5. Provides detailed output showing: - Removed: dead agents - Cleaned: orphaned resources - Created: core agents and workspaces Test Coverage: - TestRepairEnsuringCoreAgents: Verifies core agents are created - TestRepairEnsuringPRShepherdInForkMode: Fork mode verification - TestRepairDoesNotDuplicateAgents: Prevents duplicates Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
…tory When running `multiclaude claude` and no session history exists, the command now generates a new session ID instead of reusing the old one. This fixes "Session ID is already in use" errors that occur when Claude exits abnormally and leaves the session ID locked. Also adds missing trigger_refresh socket command documentation. Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add support for machine-readable JSON output of the command tree: - `multiclaude --json` outputs full command tree as JSON - `multiclaude --help --json` same as above - `multiclaude <cmd> --json` outputs that command's schema This enables LLMs and automation tools to programmatically discover available commands, their descriptions, usage patterns, and subcommands. Includes CommandSchema struct for clean JSON serialization that filters internal commands (prefixed with _) from output. Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
…nflicts The 'multiclaude claude' command was failing with "Session ID already in use" error when Claude was already running in the agent context. This happened because the command would attempt to restart Claude with --session-id or --resume flags without checking if a Claude process was already active with that session ID. Changes: - Add process alive check for stored agent PID before restarting - Add double-check for any running process in the tmux pane - Provide helpful error messages with steps to exit and restart - Import syscall package for signal-based process detection The fix detects: 1. If the stored agent PID is still running 2. If a different process is running in the tmux pane Users now get clear instructions on how to properly restart Claude or attach to the existing session. Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Changed two error messages to start with lowercase to comply with Go's error formatting conventions (staticcheck rule ST1005): - "Claude is already running..." → "claude is already running..." - "A process..." → "a process..." These are multi-line user-facing error messages that remain helpful while following Go's convention that error strings should not be capitalized (since they may be wrapped in other error contexts). Fixes lint failures in PR dlorenc#321 Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Restructures the CLI help output to be more user-friendly: - Add QUICK START section showing the 4 most common commands - Group commands into 6 categories: DAEMON, REPOSITORIES, AGENTS, COMMUNICATION, MAINTENANCE, META - Hide 7 redundant aliases from help (still functional, just not displayed) - Add Hidden and Category fields to Command struct for flexibility The help output now shows 21 focused commands instead of 28 scattered entries, making it much easier for new users to understand what to do. Also adds CLI_RESTRUCTURE_PROPOSAL.md documenting the analysis and future restructuring options for v2.0. Aligns with ROADMAP P2 "Better onboarding". Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add LongDescription field to Command struct for detailed help text - Add comprehensive help for 'repo hibernate' explaining token consumption - Show specific active agents in 'multiclaude status' (supervisor, merge-queue, etc.) instead of just 'X core, Y workers' - Add token consumption warning when agents are active - Point users to 'hibernate --all' to stop token usage This makes hibernate more discoverable and helps users understand that running agents continuously consume API tokens. Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
When run inside an agent worktree, `multiclaude refresh` now syncs just that worktree directly instead of triggering a global refresh via daemon. This gives agents immediate feedback and control over their sync process. Changes: - Add context detection: automatically identifies agent worktree from cwd - Add direct refresh: syncs single worktree using worktree.RefreshWorktree() - Add --all flag: explicitly triggers global refresh (previous behavior) - Update /refresh slash command to recommend CLI method - Provide detailed output: fetch status, rebase info, conflict handling Behavior: - Inside agent worktree: refreshes that worktree directly with feedback - Outside agent context: triggers daemon-based global refresh - With --all flag: always triggers daemon-based global refresh P0 Roadmap item: Worktree sync Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
P0 Roadmap item: Clear error messages This change ensures every user-facing error tells users: 1. What went wrong (clear, categorized message) 2. How to fix it (actionable suggestion) Changes: - Add 17 new error constructors to internal/errors for common failure modes - Update ~50 error handling locations in CLI to use structured errors - Replace raw fmt.Errorf calls with CLIError providing suggestions New error types added: - RepoAlreadyExists, DirectoryAlreadyExists, WorkspaceAlreadyExists - InvalidWorkspaceName, InvalidTmuxSessionName - LogFileNotFound, InvalidDuration, NoDefaultRepo - StateLoadFailed, SessionIDGenerationFailed, PromptWriteFailed - ClaudeStartFailed, AgentRegistrationFailed - WorktreeCleanupNeeded, TmuxWindowCleanupNeeded, TmuxSessionCleanupNeeded - WorkerNotFound, AgentNoSessionID Before: Error: failed to register worker: connection refused After: Error: failed to register worker with daemon: connection refused Try: multiclaude daemon status Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Workers now receive guidance to use /sc:index-repo for large codebase exploration, achieving 94% token reduction (58K → 3K tokens). This improves memory usage and search efficiency for complex tasks. The guidance is embedded in the worker template and will be included in system prompts for all new worker agents. Task: If a skill is provided such as QMD that enables better memory and tokens and searching, make sure that Multiclaude uses it. Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
## Problem Messages were piling up in the filesystem because acknowledged messages were never deleted. The messages.Manager had a DeleteAcked() method, but it was never called by the daemon's message routing loop. Evidence: - 128 message files accumulated in production - 17 acked messages that should have been deleted - 111 delivered messages never acknowledged ## Root Cause The messageRouterLoop delivered messages and marked them as "delivered", but had no cleanup mechanism. The DeleteAcked() method existed but was only used in tests. ## Solution Added automatic cleanup of acknowledged messages to the routeMessages() function. After delivering pending messages to each agent, the loop now calls DeleteAcked() to remove any messages that have been acknowledged. The cleanup: - Runs every 2 minutes as part of the normal message routing cycle - Only deletes messages with status "acked" - Logs cleanup activity at debug level for visibility - Handles errors gracefully without disrupting message delivery ## Testing - Added TestMessageRoutingCleansUpAckedMessages to verify cleanup works - All existing daemon tests pass - Verified in production: 17 acked messages cleaned up after daemon restart ## Impact - Prevents unbounded growth of message files - Reduces filesystem clutter - Makes the message system more reliable - No breaking changes to message API or behavior Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
…c#333 on pr-triage-b2 Tests from multiclaude workers (silly-otter, lively-otter, clever-bear): - PR dlorenc#338: Token-aware status display, hibernate help, rich list_repos response - PR dlorenc#339: Context-aware refresh, worktree path detection, --all flag parsing - PR dlorenc#337: Categorized help (worker report captured, tests via CLI assertions) - PR dlorenc#333: Enhanced repair (worker report captured, daemon handler tests) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The "Token Efficiency" section referenced a non-existent sc:index-repo slash command with fabricated claims (94% token reduction, 58K->3K tokens). This confused workers into trying to run a command that doesn't exist. Introduced by PR dlorenc#341, the skill was likely hallucinated by the agent. Also investigated the reported format.Dimmed bug in systemStatus (cli.go:1081) - confirmed it is NOT a bug. format.Dimmed is a void function that already prints via Dim.Printf, so the code works correctly. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This draft proposal is obsolete — key recommendations have already been implemented: - Categorized help output (PR dlorenc#337) - Hidden command field concept in Command struct - agents command restructuring The remaining suggestions (deprecation warnings, guide command, templates rename) can be revisited as standalone issues if needed. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…lRepair Previously, localRepair logged errors from ensureCoreAgents and ensureDefaultWorkspace as warnings (only visible with --verbose) but returned nil, implying success. Callers had no way to know repair had issues. Now errors are accumulated in a slice and returned as a combined error message so callers can detect partial repair failures. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…faultWorkspace, spawnCoreAgent These functions (from PR dlorenc#333) had zero test coverage. Adds 8 tests covering: - ensureCoreAgents with empty state, missing tmux session, existing agents, fork mode - ensureDefaultWorkspace with empty state, existing workspace, missing tmux session - spawnCoreAgent error message includes resp.Error details Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Extract shared core agent decision logic into state.MissingCoreAgents() to eliminate ~60 lines of duplicated fork-mode detection and config defaulting. Fix daemon ensureDefaultWorkspace to use writePromptFile() helper and claudeRunner.Start() instead of raw os.WriteFile/exec.Command, making it consistent with all other agent creation paths. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
… to restart handleRestartAgent requires the agent to already exist in state, causing a chicken-and-egg failure when repair tries to create missing core agents like merge-queue. Now spawnCoreAgent does full creation: tmux window, prompt file, Claude process, and state registration. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Resolved conflicts with dlorenc#337 categorized help: kept categorized output and added JSON hint line. showCommandHelp now supports both outputJSON parameter and root command redirect to categorized help. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…nc#336, dlorenc#340, dlorenc#342 Add 659 lines of tests covering: - All 18 structured error constructors from PR dlorenc#340 (individual + bulk format test) - JSON CLI output edge cases from PR dlorenc#335 (empty/nested/all-internal subcommands) - Structured CLIError validation for workspace names from PR dlorenc#340 integration - Message routing edge cases from PR dlorenc#342 (no acked, mixed ack status) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Mark completed items (Hidden field, categorized help, aliases hidden) and preserve remaining proposals (guide command, agents→templates rename, deprecation warnings, v2.0 alias removal). Trimmed redundant sections. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…s, fixes, and enhancements PR triage: merge 12 upstream PRs that passed CI
Author
CI StatusThis PR is from a fork. GitHub Actions requires maintainer approval before CI workflows will run on fork PRs. Local verification (2026-03-13):
Ready for maintainer review once CI workflows are approved. |
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
2 participants
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.
Summary
Batch integration of 12 upstream PRs from
dlorenc/multiclaudewith additional test coverage, bug fixes, and enhancements. Cross-reviewed adversarially using Gemini Pro & Claude Opus 4.6.Upstream PRs Included
Additional Fixes (not in upstream PRs)
Test plan
go test ./...passesgo test ./test/...E2E tests passmake check-allpasses (full CI equivalent)🤖 Generated with Claude Code