Digest / August 1, 2025
Claude Code Issues Digest — August 1, 2025
63 new issues filed ·
19 resolved in anthropics/claude-code (UTC day).
New issues (63)
- [BUG] Haiku agents do not work if MAX_THINKING_TOKENS is set.
- Anthropic API Context Limit Exceeded During PDF Compaction
- [BUG] --settings flag not work
- Conversation Compaction Failure: Unexpected Error When Attempting to Reduce Context Length
- Request: Model Indicator in the UI
- Desktop Commander MCP: Performance degradation due to infinite React render loop in skills system
- Claude Desktop/Web: Severe UI performance regression due to infinite React render loop in skills system
- [BUG] Console output on Windows is overwritten during file edits, making conversation history unreadable
- VSCode IDE Connection: Multiple Duplicate Workspace Instances Detected
- [BUG] Cursor incorrectly positioned when pasting multi-line code
- [BUG] No agents or commands available in 1.0.65
- MacOS German Keyboard: Cannot Enter "@" Symbol in Jetbrains Plugin
- Feature Request: Add 'prompt' list to permission rules for conditional confirmation
- Feature: Centralize and Expose File Filtering Configuration in `settings.json`
- Feature Request: Declarative Conditional Permissions Based on Tool Outcomes
- Feature: Declarative Tool Filtering in MCP Server Configuration
- Feature Request: Dynamic Sourcing of Allowed Bash Commands from Project Files
- Feature Request: Scoped Context Passing for Subagents
- Diagnostic Command Fails with Empty Response
- Invalid Model Name for Haiku Agent in Command Execution
- [BUG] subagents too slow and cost more tokens
- Unintended File Deletion During Code Update Process
- Custom Command Discovery Failure and Hidden Directory Reference Limitation
- [BUG] intermittent sporadic usage policy violation
- [BUG] 任务执行过程中，自己断了
- [BUG] IDE detection incorrectly warns when IDE is open in subdirectory of workspace
- Performance Degradation: Slow Task Execution on Max 20x Plan
- Missing Input Cursor in Bug Submission Interface
- URL Generation Fails Due to Excessive Length on GitHub Issue Submission
- Anthropic API Policy Violation Detection Incorrectly Blocking All User Inputs
- Concurrent Sub-Agent Execution Causing Intermittent Hallucinations
- External Editor Support to compose prompt in Claude Code
- Anthropic API: False Positive Usage Policy Violation Error
- [BUG] Weird bug in Claude Code WSL when reading a PDF file
- Claude Code MCP Execution Failure Despite Reported Success
- claude --resume creates a new session
- [BUG] `auto-accept` is not working
- [BUG] file named nul created on windows
- OTEL AWS sigv4 signature support
- [BUG]: CLI error with system prompt and doesn't recognize permissions
- [BUG] Subagent Model Selection not compatible with LLM Proxy with custom/short model names
- Invalid JSON Parsing Error in API Request Body
- File Selection Fails: "@" Trigger Only Shows Agents, Not Project-Specific Files
- [BUG]
- Code Generation Instability: Incomplete Task Execution and Partial Code Truncation
- [BUG] WebAssembly memory allocation error with --dangerously-skip-permissions after kernel update
- Feature Request: Add model selection support for custom commands (similar to subagents)
- Multiple mcpServers sections in .claude.json silently override each other and project-specific .mcp.json files are not loaded
- Unable to List Directories in Claude Code CLI
- KEGG-MCP Server: Connection or Access Failure
- Recursive Call Stack Overflow in Text Manipulation Method
- Feature Request: Add support for agent to agent communication
- Invalid JSON Request: Malformed Surrogate Character in API Payload
- [BUG]
- [BUG] Agents and commands broken on nixos
- Interrupt functionality not working
- Unreadable Text Color in Light Theme for /cost and /model Commands
- RFE: add retry support to Fetch tool
- Feature Request: Support for Shell-Style / Emacs-Inspired Keybindings in Message Input
- Anthropic API Request Exceeds Context Limit Maximum Token Threshold
- Invalid JSON Parsing Error in API Request Body
- Claude Code Memory Leak - Process Grows to 120+ GB RAM and Gets OOM Killed
- Claude defaults to training data patterns instead of following CLAUDE.md project guidelines
Resolved issues (19)
- Desktop Commander MCP: Performance degradation due to infinite React render loop in skills system
- Claude Desktop/Web: Severe UI performance regression due to infinite React render loop in skills system
- [BUG] No agents or commands available in 1.0.65
- [BUG] 任务执行过程中，自己断了
- [BUG] MCP server "supabase" Server stderr: /usr/bin/bash: Files\Git\bin\bash.exe: No such file or directory
- Missing Input Cursor in Bug Submission Interface
- URL Generation Fails Due to Excessive Length on GitHub Issue Submission
- [BUG] Permission deny patterns not working for Read/Write tools in settings.local.json
- [BUG]
- [BUG] Permission settings unclear/inconsistent - security issues
- [BUG] Bash:* permission works inconsistently - randomly prompts for some commands
- [Feature] Model Selection of Sub Agents
- Multiple mcpServers sections in .claude.json silently override each other and project-specific .mcp.json files are not loaded
- Interrupt functionality not working
- Claude Code IDE integration opens a diff window for accepting/declining suggested changes
- Model Configuration Support for Sub Agents
- Unable to paste images into Claude Code TUI interface
- [BUG] Subagent Model Selection not compatible with LLM Proxy with custom/short model names
- External Editor Support to compose prompt in Claude Code