Claude Code
島 11 · 卡片 248
☰
目錄
登入
Moss
Cairn
今天想往哪走？
搜尋：Claude Code
地圖
Claude Code
島 11 · 卡片 248
☰
目錄
Claude Code · 島嶼
主題 0
目錄
跟地圖是同一份內容，只是換成清單。
Configuration & Memory Files
卡片 30
settings.json layering: file locations, precedence, and permissions merging
13
›
AGENTS.md compatibility (cross-agent shared instructions)
8
›
CLAUDE.md rule drift in long sessions (and when to use hooks instead)
5
›
Output styles: format, scope, and interaction with CLAUDE.md / system prompt
2
›
Permissions model: allow/ask/deny priority, defaultMode, and tool specifier syntax
2
›
MCP Servers & Tools
卡片 56
Adding an MCP server: transports, scopes, and the claude mcp add command
13
›
MCP server shows Connected but exposes 0 tools
10
›
Custom model endpoints and gateway routing
6
›
Custom slash command argument handling and frontend stripping bugs
7
›
Hook Events and Limitations
11
›
PreToolUse:Bash hooks silently not firing (v2.1.176+ regression)
8
›
MCP server trust audit and supply-chain vetting
1
›
Runtime Behavior & Recovery
卡片 85
CLAUDE.md rule drift in long sessions (and when to use hooks instead)
5
›
PreToolUse:Bash hooks silently not firing (v2.1.176+ regression)
8
›
Context window overflow and /compact failure modes
16
›
API 400 'tool_use without tool_result' and parallel-tool conversation corruption
4
›
Plan mode executing destructive commands without ExitPlanMode (v2.1.83+ regression)
3
›
Subagent recursive self-spawn and orchestration explosions
10
›
Subagent Write/Bash silently not persisting to disk
3
›
settings.json layering: file locations, precedence, and permissions merging
13
›
Hook Events and Limitations
11
›
Plan mode as architectural choice: prompt-level vs tool-level enforcement
3
›
Session rewind safety and destructive defaults
2
›
Extended thinking budget and interleaved thinking crashes
1
›
Sandbox startup race and network timeouts
1
›
devcontainer firewall: init-firewall.sh overwrite and stale allowlist
1
›
Worktree lifecycle management: Windows EEXIST and killed-agent stale locks
3
›
Terminal scrollback stroboscope and viewport reset in long sessions
2
›
CI/CD with claude-code-action
卡片 60
Authenticating claude-code-action: API key vs Max/Pro subscription
12
›
OIDC / token-exchange failures and bot-triggered workflow 401s
7
›
CI install hangs and SDK crashes in claude-code-action
9
›
CI permissions syntax: 'Bash' vs 'Bash(*)' and the sticky-review comment
9
›
Windows native binary stability and crash recovery
3
›
Session vs weekly quotas: why usage feels wrong
19
›
CI self-hosted runner state collision and idempotent marketplace install
1
›
Usage, Quotas & Cost Control
卡片 79
Context window overflow and /compact failure modes
16
›
Subagent recursive self-spawn and orchestration explosions
10
›
Session vs weekly quotas: why usage feels wrong
19
›
Token-spend levers: the cost-control playbook
21
›
Hook Events and Limitations
11
›
Subagent tool scoping and context waste
1
›
Headless `-p` mode and background process lifecycle
1
›
Terminal & Environment Integration
卡片 8
Windows native binary stability and crash recovery
3
›
Image paste in WSL2 / Windows Terminal: Ctrl+V unreliable, use drag-drop or file paths
3
›
Windows nul device file artifacts: creation, deletion, and prevention
1
›
Windows shell selection via shellPath: Git Bash auto-detection causing find/ls/grep hangs
1
›
IDE & Plugin Integration
卡片 12
JetBrains plugin IDE integration failure in WSL: WSLENV variable passthrough
5
›
Image paste in WSL2 / Windows Terminal: Ctrl+V unreliable, use drag-drop or file paths
3
›
Output styles: format, scope, and interaction with CLAUDE.md / system prompt
2
›
VS Code extension session persistence across folder switches
1
›
Claude Desktop / Claude Code missing native WSL remote integration
1
›
Model Quality & Performance Issues
卡片 4
Model quality regression: official postmortem on context-routing, output corruption, and token-selection bugs
4
›
Subscription Credentials & Third-Party CLI Integration
卡片 5
Pro/Max credentials via third-party CLI: policy, abuse classifier risks, and when to use API key instead
3
›
Account Registration & SMS Verification Failures
1
›
Organization Disabled: API key conflict vs account suspension
1
›
Session Architecture & Workflow
卡片 20
Session initialization and lifecycle: explore → plan → verify → iterate
5
›
Configuration layering: CLAUDE.md vs CLAUDE.local.md vs skills vs subagents
1
›
Subagent definition schema and scope resolution
1
›
Hatch / Hidden venv discovery and per-invocation environment setup
1
›
Subagent recursive self-spawn and orchestration explosions
10
›
Agent loop architecture and harness design patterns
1
›
Skill invocation control matrix and content lifecycle
1
›
Agent Orchestration & Multi-Session Workflows
卡片 1
Multi-agent orchestrator tool types and selection matrix
1
›
問 Moss、貼網址或記下觀察…