Conversation
Adds support for displaying the build version via: - `multiclaude --version` or `multiclaude -v` flags - `multiclaude version` subcommand The version can be set at build time via ldflags: go build -ldflags "-X 'github.com/dlorenc/multiclaude/internal/cli.Version=$(git rev-parse HEAD)'" Defaults to "dev" when not set. Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Enhances the version command with: - Semver-formatted version strings (0.0.0+commit-dev for dev builds) - GetVersion() helper that uses VCS info from Go build - IsDevVersion() helper for checking development builds - JSON output support (--json flag) for machine-readable format JSON output includes: - version: semver-formatted version string - isDev: boolean indicating if this is a dev build - rawVersion: the raw Version variable value For dev builds, GetVersion() extracts the commit hash from VCS build info embedded by Go and returns "0.0.0+<commit>-dev". For release builds (Version set via ldflags), returns Version as-is. Reference: https://semver.org Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
…n-command Upstream contrib/version command
State package: - Fix race condition in concurrent saves by using unique temp files via os.CreateTemp instead of a fixed .tmp filename - Add proper cleanup of temp files when write/rename errors occur - Fix GetAllRepos() to include TaskHistory in the deep copy Worktree package: - Fix HasUnpushedCommits to properly detect and error on non-git dirs instead of silently returning false - Add CleanupOrphanedWithDetails to report removal errors instead of silently ignoring them, while keeping backwards compatibility Add comprehensive tests for: - Concurrent save operations - TaskHistory deep copy verification - Temp file cleanup verification - UpdateAgentPID and UpdateTaskHistorySummary functions - Non-git directory error handling - CleanupOrphanedWithDetails functionality Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Fix pkg/tmux doc.go example to include context.Context parameters - Add IsBinaryAvailable() method to pkg/claude Runner for CLI availability check - Document Config.Resume behavior with usage guidance - Add comprehensive context cancellation tests for tmux operations - Add error path tests for prompt WriteToFile and Loader Test coverage improvements: - pkg/tmux: 82.3% → 89.6% - pkg/claude: 88.2% → 90.0% - pkg/claude/prompt: 90.9% → 95.5% Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
This PR addresses code quality issues in the internal/cli package: 1. Remove dead code: Deleted unused SelectFromListWithDefault function from selector.go 2. Reduce duplication: Extracted formatAgentStatusCell helper function to replace 3 identical switch statements for formatting status cells with colors (in listWorkers, listWorkspaces, and workspace display) 3. Add analysis document: Created docs/cli-refactoring-analysis.md with detailed findings and recommendations for future improvements The cli.go file is 5,052 lines with 28.3% test coverage. The analysis document outlines a phased approach for further improvements: - Phase 1: Quick wins (completed in this PR) - Phase 2: File splitting into logical files - Phase 3: Test coverage improvement - Phase 4: Extract common patterns Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Extract common agent startup code into startAgentWithConfig() - Unify writePromptFile and writePromptFileWithPrefix functions - Remove ~44 lines of duplicated code from daemon.go - Add comprehensive tests for: - getRequiredStringArg helper function - recordTaskHistory function (success/failure cases) - linkGlobalCredentials function - repairCredentials function - isLogFile edge cases These changes improve code maintainability by reducing duplication in the agent startup code path and increase test coverage for previously untested utility functions. Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
fix: Improve robustness of state and worktree packages
docs: Improve public pkg documentation and test coverage
refactor: CLI package code quality improvements
refactor: Consolidate agent startup logic and add daemon tests
…val logic (dlorenc#231) This commit includes small, safe refactoring improvements that enhance code readability and maintainability without changing behavior: 1. Apply gofmt -s simplifications: - Simplify struct field alignment in cli.go, state.go, worktree.go - Remove redundant spacing in inline comments (50 lines) 2. Extract duplicated directory removal code: - Add removeDirectoryIfExists() helper function - Replace 5 duplicate code blocks in stopAll() command - Reduces code duplication by ~35 lines All tests pass. No behavior changes. Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com>
Add a new CI job that runs 'gofmt -s -l .' to ensure all Go code is properly formatted before merge. Also fix formatting in 3 existing test files that had minor whitespace issues. Co-authored-by: Test User <test@example.com> Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
This is PR 1 of issue dlorenc#233 (Configurable agents via markdown definitions). Changes: - Create internal/templates/agent-templates/ directory with current prompts extracted to markdown files: - merge-queue.md - worker.md - reviewer.md - Add internal/templates package that embeds templates and provides CopyAgentTemplates() function - Add RepoAgentsDir() method to pkg/config for ~/.multiclaude/repos/<repo>/agents/ - Update initRepo to copy templates to per-repo agents directory on init No behavior changes yet - daemon still reads from old locations (internal/prompts). This is preparation for making agents user-configurable. Related: dlorenc#233 Co-authored-by: Test User <test@example.com> Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
…enc#237) This PR implements the infrastructure for reading and managing agent definitions from markdown files, as specified in issue dlorenc#233. Changes: - Add new internal/agents package with: - Reader to read agent definitions from local and repo directories - Definition struct to represent parsed agent definitions - MergeDefinitions function implementing resolution order where checked-in repo definitions win over local on filename conflict - ParseTitle and ParseDescription helpers to extract metadata from markdown - Add 'multiclaude agents list' command that shows available agent definitions for a repository, including source (local vs repo) - Add comprehensive tests for the new functionality Agent definitions are read from: 1. ~/.multiclaude/repos/<repo>/agents/*.md (local user definitions) 2. <repo>/.multiclaude/agents/*.md (checked-in team definitions) When the same filename exists in both locations, the checked-in repo definition takes precedence. Closes dlorenc#233 (PR 2 of 3) Co-authored-by: Test User <test@example.com> Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
…nc#233 PR 3) (dlorenc#238) This PR implements the third phase of issue dlorenc#233 (Configurable agents via markdown definitions), making the supervisor the orchestrator for agent spawning. Changes: - Add ParseClass() and ParseSpawnOnInit() methods to agents.Definition for extracting agent metadata from markdown content - Add spawn_agent daemon handler that accepts inline prompts (name, class, prompt text) instead of hardcoded agent types - Update daemon startup to read agent definitions and send them to the supervisor on initialization - Update supervisor prompt with new "Agent Orchestration" section explaining how to interpret and use agent definitions The daemon now: 1. Reads agent definitions from ~/.multiclaude/repos/<repo>/agents/ and <repo>/.multiclaude/agents/ using the agents.Reader 2. Sends all definitions to the supervisor via a message on startup 3. Provides a spawn_agent socket command for spawning agents with custom prompts The supervisor now understands: - How to receive and interpret agent definitions - Agent classes (persistent vs ephemeral) - Spawn conditions from markdown content This is part of the transition to make agents user-configurable instead of hardcoded. Future work will give the supervisor more direct control over when and how to spawn agents. Co-authored-by: Test User <test@example.com> Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
- Add AgentTypeGenericPersistent constant for persistent agents spawned from definitions, avoiding semantic confusion with AgentTypeMergeQueue - Update daemon health check, wake loop, and restart logic to recognize the new agent type for auto-restart behavior - Simplify agent definition parsing: remove keyword/frontmatter detection and let the supervisor (Claude) interpret raw markdown definitions - Update supervisor prompt with clearer instructions for interpreting agent definitions and spawning agents The keyword-based parsing was fragile and could have false positives. Moving interpretation to Claude makes the system more flexible and reduces brittle parsing code. Co-authored-by: Test User <test@example.com> Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
…PR 4) (dlorenc#243) This PR completes the configurable agents implementation by fixing the remaining gaps between supervisor orchestration and daemon agent spawning: 1. Add `multiclaude agents spawn` CLI command - Allows spawning agents from prompt files - Connects supervisor orchestration with daemon's spawn_agent handler - Usage: multiclaude agents spawn --name <n> --class <c> --prompt-file <f> 2. Add `multiclaude agents reset` CLI command - Resets agent definitions to defaults by re-copying from templates - Useful for recovering from customization issues 3. Remove direct merge-queue startup from daemon - Merge-queue is now spawned by supervisor based on definitions - Daemon sends merge-queue config (enabled/track-mode) to supervisor - Supervisor decides whether and how to spawn merge-queue 4. Remove duplicate prompts from internal/prompts/ - Removed merge-queue.md, worker.md, review.md (now in agent-templates) - Keep only supervisor.md and workspace.md (hardcoded agents) - Update prompts.go and tests for new behavior - CLI functions now read worker/merge-queue prompts from agent definitions 5. Update supervisor prompt - Use new `multiclaude agents spawn` command instead of message-based approach - Document merge-queue configuration section in definitions message Co-authored-by: Test User <test@example.com> Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
…renc#240) * refactor: Extract branch cleanup helper to eliminate duplication Extract cleanupOrphanedBranchesWithPrefix() helper function to remove 45 lines of duplicated code in localCleanup(). The same logic was repeated for work/* and workspace/* branch cleanup. Changes: - Add cleanupOrphanedBranchesWithPrefix() helper (38 lines) - Replace duplicated blocks with helper calls (reduces 45 lines to 7) - Net reduction: ~38 lines of code - No behavior changes Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com> * refactor: Extract common periodic loop pattern in daemon Extract periodicLoop() helper to eliminate repetitive loop boilerplate across three daemon background loops (health check, message router, wake). Changes: - Add periodicLoop() helper function with configurable startup/tick callbacks - Replace healthCheckLoop() body with periodicLoop call - Replace messageRouterLoop() body with periodicLoop call - Replace wakeLoop() body with periodicLoop call - Net reduction: 18 lines of duplicated code Benefits: - Reduces duplication of ticker/select/context pattern - Centralizes loop cancellation logic - Makes loop behavior more consistent and testable - Simplifies adding new background loops in the future No behavior changes - identical functionality preserved. Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com>
* refactor: Use standard repo resolution in addWorkspace Replace inline repository resolution logic with call to resolveRepo() to ensure consistent behavior across all commands. Changes: - Remove manual repo resolution (11 lines) - Use c.resolveRepo(flags) instead (3 lines) - Net reduction: 8 lines Benefits: - Adds git remote URL matching (was missing) - Adds daemon current-repo query (was missing) - Consistent error messages across commands - Better user experience in git repositories Previously, addWorkspace() only checked --repo flag and cwd inference, while other commands also checked git remotes and daemon default repo. This inconsistency meant commands like 'work' would find a repo but 'workspace add' would fail in the same directory. No behavior changes for existing working cases - only adds fallback options that were previously unavailable. Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com> * test: Fix flaky tmux tests in CI by adding robust session check Add skipIfTmuxCantCreateSessions() helper that verifies tmux can actually create sessions, not just that the binary exists. In CI environments without a proper terminal, tmux may be installed but unable to create sessions. The old IsTmuxAvailable() only checked if `tmux -V` works. Updated 10 tests to use the new helper for consistent behavior: - TestHealthCheckLoopWithRealTmux - TestHealthCheckCleansUpMarkedAgents - TestMessageRoutingWithRealTmux - TestWakeLoopUpdatesNudgeTime - TestWakeLoopSkipsRecentlyNudgedAgents - TestRestoreTrackedReposExistingSession - TestRestoreDeadAgentsWithExistingSession - TestRestoreDeadAgentsSkipsAliveProcesses - TestRestoreDeadAgentsSkipsTransientAgents - TestRestoreDeadAgentsIncludesWorkspace - TestHealthCheckAttemptsRestorationBeforeCleanup Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com> Co-authored-by: Test User <test@example.com>
Add comprehensive tests to improve coverage, focusing on critical business logic and error handling paths. Achieved 100% coverage for internal/format package and significant improvements in other packages. Coverage improvements: - internal/format: 78.4% → 100.0% (+21.6%) - internal/prompts/commands: 76.2% → 85.7% (+9.5%) - internal/daemon: 59.2% → 59.7% (+0.5%) - internal/cli: 29.1% → 30.1% (+1.0%) Added tests cover: - Format package: Header, Dimmed, ColoredTable.Print, totalWidth calculations - Prompts/commands: Error handling for directory generation and setup - Daemon: Wait, trigger functions (health check, message routing, wake, worktree refresh) - CLI: Version commands, help, execute with edge cases All new tests follow existing patterns and use proper test helpers. Added COVERAGE_IMPROVEMENTS.md documenting changes and recommendations. Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com>
…lorenc#244) - Fix bug where TypeMergeQueue mapped to wrong filename "REVIEWER.md" instead of "MERGE-QUEUE.md" - Add comprehensive tests for handleSpawnAgent daemon handler covering missing args, invalid class, repo not found, and agent exists cases - Add tests for spawnAgentFromFile CLI command covering validation errors and non-existent prompt file - Add tests for resetAgentDefinitions CLI command covering fresh creation and reset scenarios Co-authored-by: Test User <test@example.com> Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
…te.go (dlorenc#246) * refactor: Extract atomic write helper to eliminate duplication in state.go Extract duplicated atomic file write logic from Save() and saveUnlocked() into a shared atomicWrite() helper function. This eliminates 36 lines of exact code duplication and ensures consistent atomic save behavior across both methods. Changes: - Add atomicWrite(path, data) helper for atomic file writes - Update Save() to use atomicWrite helper - Update saveUnlocked() to use atomicWrite helper - All 45 state package tests pass Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com> * refactor: Extract daemon request helper to reduce duplication in CLI Add sendDaemonRequest() helper method to centralize socket client creation and error handling. This eliminates repetitive boilerplate code across CLI commands and ensures consistent error handling. Changes: - Add sendDaemonRequest(command, args) helper to CLI struct - Update 6 functions to use the helper: - stopDaemon - listRepos - setCurrentRepo - getCurrentRepo - clearCurrentRepo - listWorkers - Net reduction: 25 lines of code eliminated - All 38 CLI tests pass Impact: - Reduces duplication across 6+ functions (with ~24 more opportunities) - Consistent error handling for daemon communication - Easier to maintain and modify daemon request behavior Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com>
…system (dlorenc#245) * refactor: Consolidate AgentType definitions and deprecate old prompt system This consolidates the parallel customization systems created during the configurable agents implementation: 1. **Consolidated AgentType definitions**: - Made state.AgentType the canonical type definition - Created deprecated aliases in prompts package for backward compatibility - Updated all call sites in cli.go and daemon.go to use state.AgentType 2. **Deprecated old LoadCustomPrompt system**: - Removed LoadCustomPrompt calls from writeWorkerPromptFile and writeMergeQueuePromptFile - Added deprecation notice to LoadCustomPrompt function - The new system (<repo>/.multiclaude/agents/) is now the single source for agent customization 3. **Updated documentation**: - AGENTS.md: Updated "Custom Prompts" section to document new system - README.md: Updated "Repository Configuration" section - SPEC.md: Updated "Role Prompts" section The old system using SUPERVISOR.md, WORKER.md, REVIEWER.md directly in .multiclaude/ is deprecated. Users should migrate to the new .multiclaude/agents/ directory structure. Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * fix: Use unique session names in skipIfTmuxCantCreateSessions The helper function was using a static session name for all tests, which could cause race conditions in CI when tests run sequentially. Now each test gets a unique session name based on the test name, and any leftover sessions are cleaned up before the check. This fixes the failing Coverage Check in CI. Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> --------- Co-authored-by: Test User <test@example.com> Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
…orenc#247) * docs: Comprehensive documentation for configurable agents feature Add documentation for the configurable agents system introduced in dlorenc#233: - README.md: Add "Agent Definitions" commands subsection with CLI reference - README.md: Add "Configurable Agents" section explaining customization - README.md: Update directory structure to include agents directories - AGENTS.md: Add CLI commands and practical example for customizing workers - docs/DIRECTORY_STRUCTURE.md: Document agent definition directories and precedence The documentation covers: - How to list, reset, and spawn custom agents - Where agent definitions are stored (local vs repo-checked) - Precedence order for agent definitions - Step-by-step example of customizing worker conventions - How to share definitions with team members Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * chore: Regenerate docs after agent definition removal Run `go generate ./pkg/config` to update DIRECTORY_STRUCTURE.md to reflect the removal of the per-repo agent definitions system. Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> --------- Co-authored-by: Test User <test@example.com> Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Add Node.js setup and Claude CLI installation to the e2e-tests job so that E2E tests can run properly instead of being skipped when the claude binary is not available. Co-authored-by: Test User <test@example.com> Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
…c#249) * test: Add comprehensive tests for DeleteRemoteBranch function Add tests covering DeleteRemoteBranch (0% -> 100% coverage) and enhance CleanupMergedBranches testing with remote deletion scenarios. Coverage improvements: - DeleteRemoteBranch: 0% -> 100% - CleanupMergedBranches: improved remote deletion path coverage - Overall worktree package: 78.6% -> 80.0% Test scenarios added: - Successfully deletes remote branch - Error handling when remote doesn't exist - Error handling when branch doesn't exist on remote - CleanupMergedBranches with deleteRemote=true Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com> * test: Improve HasUnpushedCommits test coverage (53.3% -> 93.3%) Add comprehensive tests for HasUnpushedCommits with tracking branches to cover the previously untested code paths for detecting unpushed commits. Coverage improvements: - HasUnpushedCommits: 53.3% -> 93.3% - Overall worktree package: 80.0% -> 81.3% Test scenarios added: - Detects unpushed commits when tracking branch exists - Handles push/pull cycle correctly - Detects multiple unpushed commits - Verifies correct behavior after pushing commits Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com> * test: Add error handling tests for templates package (75% -> 82.1%) Add comprehensive error handling and edge case tests for the templates package, focusing on filesystem operations and error scenarios. Coverage improvements: - CopyAgentTemplates: 70.0% -> 80.0% - Overall templates package: 75.0% -> 82.1% Test scenarios added: - Error handling for read-only destinations - Nested directory creation - Edge cases with empty/current directory paths - Consistency verification between ListAgentTemplates and CopyAgentTemplates Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com> * test: Add error handling tests for messages package (82.2% -> 91.1%) Add comprehensive error handling tests for the messages package, covering write/read failures, permission errors, and edge cases. Coverage improvements: - Send: 75.0% -> 100.0% - write: 66.7% -> 77.8% - read: 87.5% -> 100.0% - CleanupOrphaned: 77.8% -> 94.4% - Overall messages package: 82.2% -> 91.1% Test scenarios added: - Send fails with invalid permissions - Get/UpdateStatus fail for non-existent messages - List/ListUnread handle non-existent directories gracefully - read handles corrupted JSON files - Delete is idempotent - CleanupOrphaned ignores files and handles non-existent repos Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com>
…r CI (dlorenc#253) The test was failing in CI environments where tmux is installed but cannot create sessions due to lack of TTY. The skipIfTmuxCantCreateSessions helper function passes because it successfully creates and destroys a test session, but subsequent session creation can still fail intermittently. This change adds a secondary skip condition at the actual session creation point, following the same pattern used in TestRouteMessageToAgent. This provides a belt-and-suspenders approach to ensure the test skips gracefully rather than failing when tmux session creation is unreliable. Fixes failing Unit Tests in PR dlorenc#248. Co-authored-by: Test User <test@example.com> Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
This commit: 1. Adds MULTICLAUDE_TEST_MODE check to daemon's startAgentWithConfig function. When test mode is enabled, skips claude binary resolution and startup while still registering agents in state. This follows the same pattern already used by the CLI code (see internal/cli/cli.go:1108). 2. Adds comprehensive daemon tests for the new test mode functionality. 3. Fixes failing coverage check by reverting the skipIfTmuxCantCreateSessions helper back to the simpler IsTmuxAvailable() check. The more complex helper was causing race conditions when creating/killing tmux sessions in CI. This fixes failing E2E tests (TestAgentsSpawnCommand, TestSpawnPersistentAgent, TestSpawnEphemeralAgent) from PR dlorenc#248 that use the daemon's spawn_agent handler. Based on PR dlorenc#252 with coverage check fix applied. Co-authored-by: Test User <test@example.com> Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
* test: Add comprehensive E2E tests for configurable agents feature Add E2E tests in test/agents_test.go for: - TestAgentTemplatesCopiedOnInit: Verifies templates are copied during init - TestAgentDefinitionMerging: Tests local + repo definition merge (repo wins) - TestAgentsListCommand: Tests the agents list command - TestAgentsResetCommand: Tests agents reset restores defaults - TestAgentsSpawnCommand: Tests spawn via daemon's spawn_agent handler - TestAgentDefinitionsSentToSupervisor: Tests definitions sent on restore - TestSpawnPersistentAgent: Tests persistent agent creation - TestSpawnEphemeralAgent: Tests ephemeral agent creation Add daemon unit tests in internal/daemon/daemon_test.go: - TestSendAgentDefinitionsToSupervisor: Tests the function that sends agent definitions to the supervisor, including: - No definitions case returns nil - Sends definitions to supervisor - Includes merge queue config when enabled - Includes disabled message when merge queue disabled - Includes spawn instructions Coverage improvement: - Daemon: 60.6% -> 62.8% Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * fix: Skip claude binary startup in daemon during test mode Add MULTICLAUDE_TEST_MODE check to startAgentWithConfig in the daemon. When test mode is enabled, the daemon skips: - Claude binary resolution - Sending the claude command to tmux - Getting the Claude process PID This fixes TestAgentsSpawnCommand, TestSpawnPersistentAgent, and TestSpawnEphemeralAgent which were failing in CI because the claude binary is not available. The pattern follows how the CLI already handles test mode. Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> --------- Co-authored-by: Test User <test@example.com> Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Remove documentation for features that don't exist: - docs/extending/EVENT_HOOKS.md - out of scope per ROADMAP - docs/extending/WEB_UI_DEVELOPMENT.md - out of scope per ROADMAP Remove one-time analysis documents: - AUDIT_REPORT.md - docs/EXTENSIBILITY.md (redundant with extending/*) - docs/EXTENSION_DOCUMENTATION_SUMMARY.md - docs/TEST_ARCHITECTURE_REVIEW.md - docs/cli-refactoring-analysis.md Simplify CLAUDE.md extensibility section and improve README with inline code examples. Total: -3,606 lines of dead/redundant documentation. Co-authored-by: Test User <test@example.com> Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
* fix: append custom agent definitions to base templates instead of replacing Custom agent definitions from .multiclaude/agents/ are now appended to the base templates from ~/.multiclaude/repos/<repo>/agents/ instead of replacing them entirely. This ensures critical instructions like 'multiclaude agent complete' are never lost when users customize their agent definitions. Changes: - Add SourceMerged constant to indicate merged definitions - Update MergeDefinitions to append content with "## Custom Instructions" separator when names match - Update unit and integration tests to verify new append behavior Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * fix: remove extra blank line for gofmt compliance Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> --------- Co-authored-by: Test User <test@example.com> Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Removed gopkg.in/yaml.v3 which was listed in go.mod but never imported. Discovered via `go mod tidy` lint warning. Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
…Ls (dlorenc#320) Fixes two bugs in the init command: 1. The init command now checks if a repository is already initialized or if the tmux session already exists before attempting to create them. Previously, users would get cryptic "exit status 1" errors when trying to re-initialize a repo. Now they get clear error messages with actionable suggestions. 2. The GitHub URL parser now supports repository names containing dots (e.g., demos.expanso.io). The regex was too restrictive and excluded dots from repo names, causing fork detection to fail for repos with dotted names. Changes: - Add pre-flight checks in initRepo to verify repo doesn't exist in state - Check if repository directory already exists before cloning - Check if tmux session exists before attempting to create it - Update regex in ParseGitHubURL to allow dots in repository names - Update test to reflect that dots in repo names are now supported Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com>
…add (dlorenc#322) Fixes state inconsistency issues when init or workspace add commands fail partway through execution. Previously, if these commands failed after creating tmux sessions/windows but before updating state, re-running would fail with cryptic tmux errors. Changes: - initRepo: Check for existing tmux session and state entry before creation - Auto-repair by killing stale tmux sessions - Clear error if repo already tracked in state - addWorkspace: Check for existing tmux window and worktree before creation - Auto-repair by killing stale tmux windows - Auto-repair by removing stale worktrees This improves P0 "Clear error messages" goal by preventing confusing error states and making the system self-healing. Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com>
Adds spec-driven proposal for unified repo lifecycle commands: - repo start: Initialize all standard agents - repo status: Comprehensive status display - repo hibernate: Pause agents, preserve state - repo wake: Resume hibernated repo - repo refresh: Sync worktrees with main - repo clean: Remove orphaned resources Includes design decisions, implementation tasks, and detailed specs. Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
…orenc#314) Addresses P0 roadmap item "Worktree sync". Adds: - trigger_refresh socket command in daemon - multiclaude refresh CLI command that triggers immediate worktree sync - Allows merge-queue or users to force refresh after PRs merge The existing 5-minute refresh loop continues unchanged; this adds on-demand capability for faster sync when needed. Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Fork package tests were failing on machines with global git URL rewrite rules (e.g., url.git@github.com:.insteadof=https://github.com/). Tests expected exact HTTPS URLs but git was returning SSH URLs due to the global insteadOf configuration. Changes: - Add gitCmdIsolated() helper that runs git with GIT_CONFIG_GLOBAL and GIT_CONFIG_SYSTEM set to /dev/null for deterministic behavior - Add urlsEquivalent() helper for semantic URL comparison (owner/repo match) instead of exact string matching - Update all test git commands to use isolated environment - Update URL comparisons to use semantic equivalence This makes tests deterministic regardless of developer's local git configuration. Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
* feat: add repo hibernate command to archive and stop work Adds `multiclaude repo hibernate` command that cleanly stops all work in a repository while preserving uncommitted changes: - Archives uncommitted changes as patch files to ~/.multiclaude/archive/<repo>/<timestamp>/ - Saves metadata (branch, task, worktree path) for each agent - Lists untracked files separately for manual restoration - Stops workers and review agents by default (--all for persistent agents) - Force-removes worktrees after archiving to ensure clean shutdown Also adds ArchiveDir to Paths config and updates all test files to include it. Usage: multiclaude repo hibernate [--repo <repo>] [--all] [--yes] Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * fix: address lint issues in hibernate command - Check error return from client.Send (errcheck) - Fix formatting with gofmt Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Implements dlorenc#305 - adds a memorable `multiclaude status` command that: - Shows daemon status (running/not running/unhealthy) - Lists tracked repos with agent counts - Gracefully handles daemon not running (no error, just shows status) - Provides helpful hints for next steps Unlike `multiclaude list` which errors when daemon is unavailable, `multiclaude status` always succeeds and shows what it can. Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
…ker (dlorenc#307) Extends worker template with two new capability sections: **Environment Hygiene:** - Shell history stealth (leading space prefix) - Pre-completion cleanup (verify no credentials leaked) **Feature Integration Tasks:** - Reuse First principle - Minimalist Extensions guidance - PR analysis workflow - Integration checklist Follows the concise style of the existing worker template. Fixes dlorenc#282 Fixes dlorenc#283 Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
* feat: Add CI guard rails for local validation Adds Makefile and pre-commit hook to run CI checks locally before pushing. This prevents the common issue where CI fails after code is already pushed, by enabling developers to run the exact same checks that GitHub CI runs. Features: - Makefile with targets matching all CI jobs (build, unit-tests, e2e-tests, verify-docs, coverage) - Pre-commit hook script for automatic validation - Updated CLAUDE.md with usage instructions Usage: make pre-commit # Fast checks before commit make check-all # Full CI validation make install-hooks # Install git pre-commit hook * Update verify-docs to fix CI * Use double quotes for regex to avoid syntax errors * Update Makefile to include verify-docs check * Fix unused verbose variable in verify-docs
Add comprehensive tests for daemon handlers with low coverage: - handleTriggerRefresh (0% → 100%) - handleRestartAgent (validation error paths) - handleSpawnAgent (argument validation and error cases) - handleRepairState (basic functionality) - handleTaskHistory (with filters and limits) - handleListAgents (with multiple agents) - handleUpdateRepoConfig (merge queue and PR shepherd settings) - handleGetRepoConfig (validation and success cases) The tests use table-driven patterns consistent with existing tests and focus on testing argument validation and error handling paths that don't require actual tmux or Claude process startup. Co-authored-by: Test User <test@example.com> Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
…c#331) Add helper functions to socket package for creating responses: - ErrorResponse(format, args...) for error responses with formatting - SuccessResponse(data) for success responses Add argument extraction helpers in daemon: - getOptionalStringArg for optional string arguments with defaults - getOptionalBoolArg for optional bool arguments with defaults Refactored all 50 handler response patterns in daemon.go to use the new helpers, improving consistency and reducing boilerplate. This also simplifies fork config parsing from ~12 lines to ~4 lines. Co-authored-by: Test User <test@example.com> Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Port changes from aronchick#66: - Add docs/TASK_MANAGEMENT.md with comprehensive guide for Claude Code's task management tools (TaskCreate/Update/List/Get) - Add internal/diagnostics package for machine-readable system diagnostics - Add 'multiclaude diagnostics' CLI command outputting JSON with: * Claude CLI version and capability detection * Task management support detection (v2.0+) * Environment variables (with sensitive value redaction) * System paths and tool versions * Daemon status and agent statistics - Add daemon startup diagnostics logging for monitoring - Update supervisor and worker prompts with task management guidance (adapted to new concise prompt style from dlorenc#302) The diagnostics endpoint helps operators understand the multiclaude environment, and task management enables agents to organize complex multi-step work while maintaining the "focused PRs" principle. Co-authored-by: Claude Opus 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>
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>
## 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>
…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>
whitmo
added a commit
that referenced
this pull request
Mar 1, 2026
Review found PR description severely misleading: claims 5 upstream PRs but contains 99 commits (full upstream sync). Code quality is mostly good but has 5 specific concerns around removed cleanup logic, fragile type inference, and removed crash notifications. Posted review comment: dlorenc#18 (comment) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Owner
Author
|
Superseded by #19 which now contains all 12 upstream PRs (batch 1 + batch 2) plus additional fixes. |
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
3 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
dlorenc/multiclaude): fix: Clean up acknowledged messages in message routing loop dlorenc/multiclaude#342, fix(cli): generate new session ID inmulticlaude claudewhen no history dlorenc/multiclaude#334, fix: lowercase error messages to satisfy staticcheck ST1005 dlorenc/multiclaude#336, feat: improve error messages with structured errors and suggestions dlorenc/multiclaude#340, feat: Add --json flag for LLM-friendly CLI help output dlorenc/multiclaude#335PRs included
--jsonflag)Test additions
internal/errors/errors_test.go— 386 lines: all 18 new error constructorsinternal/cli/cli_test.go— 142 lines: JSON output edge cases, structured workspace validationinternal/daemon/daemon_test.go— 131 lines: message routing edge casesTest plan
go test ./...passes (verified locally)go build ./cmd/multiclaudesucceeds (verified locally)🤖 Generated with Claude Code