fix: Windows command line length limit errors #7
Conversation
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
On Windows, command line arguments are limited to ~8KB. This caused "The command line is too long" errors when: 1. client.py passed **os.environ to MCP server config (PATH alone can exceed 2KB on Windows) 2. spec_chat_session.py passed 18KB skill content as system_prompt Changes: - client.py: Remove env from MCP config, pass --project-dir as CLI arg - mcp_server/feature_mcp.py: Add argparse support for --project-dir (with fallback to PROJECT_DIR env var for backward compatibility) - spec_chat_session.py: Use short system_prompt, pass skill content in first message instead Tested on Windows 10 with long PATH environment variable.
- Use short system_prompt to avoid Windows command line limits - Pass project context in first message instead (like spec_chat_session) - Use absolute path for MCP server script to fix module resolution - Add ROOT_DIR constant to client.py for consistent path handling The long system_prompt (6331 chars) was causing Claude CLI initialization to timeout on Windows. This follows the same pattern as spec_chat_session.py which uses a short system_prompt and passes content in the first message. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Implements a coordinator-worker pattern for running multiple agents in parallel: - Add FeatureStatus enum (pending/in_progress/passing/conflict/failed) - Add atomic feature claiming via SQLite CTE with RETURNING - Add lease-based heartbeat for stale claim recovery (30min timeout) - Add WorktreeManager for git worktree lifecycle management - Add ParallelCoordinator for spawning/monitoring workers - Add parallel_agent.py CLI entry point (--workers 1-10) - Update process_manager to choose single vs parallel agent script - Add UI worker count selector (cycles 1→3→5) - Add ruff.toml config and fix lint errors Usage: CLI: python parallel_agent.py --project-dir my-app --workers 5 UI: Click users icon to select worker count before starting 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Critical fixes: - Add init_database_direct() for DB initialization without MCP server - Call init_database_direct() in coordinator before any MCP tool calls - Add --worktree-path and --feature-id CLI args to autonomous_agent_demo - Add work_dir parameter to separate code changes from DB access - Add feature_get_by_id MCP tool for fetching specific features - Add get_coding_prompt_parallel() with override instructions for bound feature - Workers now use feature_get_by_id instead of feature_get_next Major fix: - Capture target branch at startup via git symbolic-ref - merge_feature_branch() now uses explicit branch checkout, not HEAD - Prevents detached HEAD issues during merge operations Minor fix: - Pass --model to worker subprocess (verified working) 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Major fixes: - Add feature_get_by_id to allowed tools in prompt templates - Copy app_spec.txt to worktrees via WORKTREE_SYNC_FILES Tests added (14 passing): - Concurrent claim uniqueness (N workers → N distinct features) - Heartbeat lease_lost behavior (wrong worker_id) - Stale reclaim (claimed_at old → returns to pending) - Merge serialization smoke test (asyncio.Lock blocks concurrent access) - Migration forwards/backwards sanity (Enum storage + passes invariant) 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- WORKTREE_SYNC_FILES now uses (src, dst) tuples to support path remapping - prompts/app_spec.txt → app_spec.txt ensures agents can read spec - Fixed misleading comment: "dest doesn't exist" not "source is newer" 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add overwrite flag to WORKTREE_SYNC_FILES tuples - prompts/app_spec.txt always overwrites root app_spec.txt (overwrite=True) - Prevents stale specs when prompts/ version is updated 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
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.
Summary
Fixes "The command line is too long" errors on Windows when running the agent or spec creation.
On Windows, command line arguments are limited to ~8KB. This caused failures when:
client.pypassed**os.environto MCP server config (PATH alone can exceed 2KB on Windows)spec_chat_session.pypassed 18KB skill content assystem_promptChanges
envfrom MCP config, pass--project-diras CLI argument insteadargparsesupport for--project-dir(with fallback toPROJECT_DIRenv var for backward compatibility)system_prompt, pass skill content in first message insteadTest plan
🤖 Generated with Claude Code