46c5862c4e
Major audit correcting misleading documentation about Claude Code behavior: ### Fixed - `--add-dir`: permissions (not context loading) - `excludePatterns` → `permissions.deny` (never existed) - `.claudeignore` removed (not an official feature) - "selective loading" myth → lazy loading reality - Invented CLI flags (`--think`, `--headless`, `--learn`) → prompt keywords - `@` file reference: "loads automatically" → "reads on-demand" ### Added - Session Search Tool (`cs`) - zero-dep bash script for finding sessions - Security section: Known limitations of permissions.deny 15 files modified, 516 insertions, 200 deletions Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
465 lines
14 KiB
YAML
465 lines
14 KiB
YAML
category: "Reference"
|
||
category_id: 10
|
||
source_file: "guide/ultimate-guide.md"
|
||
|
||
questions:
|
||
- id: "10-001"
|
||
difficulty: "junior"
|
||
profiles: ["junior", "senior", "power"]
|
||
question: "What keyboard shortcut enters Plan Mode (toggles plan/execute)?"
|
||
options:
|
||
a: "Ctrl+P"
|
||
b: "Shift+Tab"
|
||
c: "Alt+P"
|
||
d: "Ctrl+Shift+P"
|
||
correct: "b"
|
||
explanation: |
|
||
`Shift+Tab` toggles between Plan Mode and Execute Mode.
|
||
|
||
Plan Mode navigation:
|
||
- `Shift+Tab` once: Toggle plan/execute
|
||
- `Shift+Tab` twice: Enter deep Plan Mode (with Opus in OpusPlan)
|
||
|
||
Plan Mode allows safe, read-only exploration before making changes.
|
||
doc_reference:
|
||
file: "guide/ultimate-guide.md"
|
||
section: "10.2 Keyboard Shortcuts"
|
||
anchor: "#session-control"
|
||
|
||
- id: "10-002"
|
||
difficulty: "junior"
|
||
profiles: ["junior", "senior", "power"]
|
||
question: "What keyboard shortcut rewinds to a previous checkpoint (undo Claude's changes)?"
|
||
options:
|
||
a: "Ctrl+Z"
|
||
b: "Esc (double-tap)"
|
||
c: "Ctrl+R"
|
||
d: "Alt+Z"
|
||
correct: "b"
|
||
explanation: |
|
||
Double-tap `Esc` (Esc×2) rewinds to the previous checkpoint.
|
||
|
||
This is equivalent to the `/rewind` command.
|
||
It undoes Claude's recent changes in the current session without creating git commits.
|
||
|
||
Use when Claude made a mistake and you want to try a different approach.
|
||
doc_reference:
|
||
file: "guide/ultimate-guide.md"
|
||
section: "10.2 Keyboard Shortcuts"
|
||
anchor: "#session-control"
|
||
|
||
- id: "10-003"
|
||
difficulty: "junior"
|
||
profiles: ["junior", "senior", "power"]
|
||
question: "What does `/compact` do?"
|
||
options:
|
||
a: "Compresses files on disk"
|
||
b: "Summarizes and compresses the conversation context"
|
||
c: "Minimizes the terminal window"
|
||
d: "Reduces Claude's response length"
|
||
correct: "b"
|
||
explanation: |
|
||
`/compact` summarizes and compresses the conversation context.
|
||
|
||
Use `/compact` when:
|
||
- Context usage reaches 70-90%
|
||
- Responses become slow
|
||
- Claude starts forgetting earlier context
|
||
|
||
This frees up context space while preserving important information.
|
||
Check context usage with `/status`.
|
||
doc_reference:
|
||
file: "guide/ultimate-guide.md"
|
||
section: "10.1 Commands Table"
|
||
anchor: "#built-in-commands"
|
||
|
||
- id: "10-004"
|
||
difficulty: "senior"
|
||
profiles: ["senior", "power"]
|
||
question: "At what context percentage should you run /compact according to best practices?"
|
||
options:
|
||
a: "50%"
|
||
b: "70-90%"
|
||
c: "95%+"
|
||
d: "Only when errors occur"
|
||
correct: "b"
|
||
explanation: |
|
||
Context management guidelines:
|
||
|
||
| Context Level | Action |
|
||
|--------------|--------|
|
||
| 0-50% | Work freely |
|
||
| 50-75% | Be selective |
|
||
| **75-90%** | **Use `/compact`** |
|
||
| 90%+ | Use `/clear` |
|
||
|
||
Proactive compaction at 70% prevents context degradation and maintains performance.
|
||
doc_reference:
|
||
file: "guide/ultimate-guide.md"
|
||
section: "10.4 Troubleshooting"
|
||
anchor: "#context-recovery"
|
||
|
||
- id: "10-005"
|
||
difficulty: "junior"
|
||
profiles: ["junior", "senior", "power"]
|
||
question: "What keyboard shortcut cancels the current operation?"
|
||
options:
|
||
a: "Esc"
|
||
b: "Ctrl+C"
|
||
c: "Ctrl+Z"
|
||
d: "Alt+C"
|
||
correct: "b"
|
||
explanation: |
|
||
`Ctrl+C` cancels the current operation.
|
||
|
||
Important shortcuts:
|
||
- `Ctrl+C`: Cancel operation
|
||
- `Ctrl+D`: Exit Claude Code
|
||
- `Ctrl+R`: Retry last operation
|
||
- `Ctrl+L`: Clear screen (keeps context)
|
||
- `Esc`: Dismiss current suggestion
|
||
doc_reference:
|
||
file: "guide/ultimate-guide.md"
|
||
section: "10.2 Keyboard Shortcuts"
|
||
anchor: "#session-control"
|
||
|
||
- id: "10-006"
|
||
difficulty: "senior"
|
||
profiles: ["senior", "power"]
|
||
question: "What is the correct permission pattern to allow ALL git commands?"
|
||
options:
|
||
a: 'Bash(git)'
|
||
b: 'Bash(git *)'
|
||
c: 'git:*'
|
||
d: 'Bash(*git*)'
|
||
correct: "b"
|
||
explanation: |
|
||
The pattern `Bash(git *)` allows any git command.
|
||
|
||
Permission pattern examples:
|
||
- `Bash(git *)` - Any git command
|
||
- `Bash(npm test)` - Exactly "npm test"
|
||
- `Edit` - All file edits
|
||
- `mcp__serena__*` - All Serena tools
|
||
|
||
Wildcards (*) enable flexible permission matching.
|
||
doc_reference:
|
||
file: "guide/ultimate-guide.md"
|
||
section: "10.3 Configuration Reference"
|
||
anchor: "#permission-patterns"
|
||
|
||
- id: "10-007"
|
||
difficulty: "junior"
|
||
profiles: ["junior", "senior", "power"]
|
||
question: "Where should project-specific CLAUDE.md be placed (committed to git)?"
|
||
options:
|
||
a: "~/.claude/CLAUDE.md"
|
||
b: "/project/.claude/CLAUDE.md"
|
||
c: "/project/CLAUDE.md"
|
||
d: "~/.config/claude/project.md"
|
||
correct: "c"
|
||
explanation: |
|
||
CLAUDE.md locations:
|
||
|
||
| Location | Scope | Committed |
|
||
|----------|-------|-----------|
|
||
| `~/.claude/CLAUDE.md` | All projects | N/A (global) |
|
||
| `/project/CLAUDE.md` | This project | **Yes** |
|
||
| `/project/.claude/CLAUDE.md` | Personal | No |
|
||
|
||
The root `CLAUDE.md` is committed and shared with the team.
|
||
The `.claude/CLAUDE.md` is personal and should be in `.gitignore`.
|
||
doc_reference:
|
||
file: "guide/ultimate-guide.md"
|
||
section: "10.3 Configuration Reference"
|
||
anchor: "#claudemd-locations"
|
||
|
||
- id: "10-008"
|
||
difficulty: "senior"
|
||
profiles: ["senior", "power"]
|
||
question: "What does the --mcp-debug flag do?"
|
||
options:
|
||
a: "Disables MCP servers"
|
||
b: "Debugs MCP server connections with verbose output"
|
||
c: "Tests MCP configurations"
|
||
d: "Enables MCP auto-discovery"
|
||
correct: "b"
|
||
explanation: |
|
||
The `--mcp-debug` flag enables debug mode for MCP server connections.
|
||
|
||
MCP debugging techniques:
|
||
```bash
|
||
claude --mcp-debug # Debug all MCP connections
|
||
/mcp # View MCP status inside Claude Code
|
||
```
|
||
|
||
Use when MCP servers aren't connecting or tools aren't appearing.
|
||
doc_reference:
|
||
file: "guide/ultimate-guide.md"
|
||
section: "10.4 Troubleshooting"
|
||
anchor: "#mcp-debugging-techniques"
|
||
|
||
- id: "10-009"
|
||
difficulty: "power"
|
||
profiles: ["power"]
|
||
question: "If an MCP server name validation fails with pattern error, what characters ARE allowed?"
|
||
options:
|
||
a: "Letters only"
|
||
b: "Letters, numbers, underscores, hyphens (max 64 chars)"
|
||
c: "Any characters except spaces"
|
||
d: "Alphanumeric and periods"
|
||
correct: "b"
|
||
explanation: |
|
||
MCP server names must match: `^[a-zA-Z0-9_-]{1,64}`
|
||
|
||
Allowed:
|
||
- Letters (a-z, A-Z)
|
||
- Numbers (0-9)
|
||
- Underscores (_)
|
||
- Hyphens (-)
|
||
- Maximum 64 characters
|
||
|
||
Not allowed:
|
||
- Spaces
|
||
- Special characters (@, #, etc.)
|
||
- More than 64 characters
|
||
doc_reference:
|
||
file: "guide/ultimate-guide.md"
|
||
section: "10.4 Troubleshooting"
|
||
anchor: "#error-1-tool-name-validation-failed"
|
||
|
||
- id: "10-010"
|
||
difficulty: "junior"
|
||
profiles: ["junior", "senior", "power"]
|
||
question: "What command shows session info including context usage and cost?"
|
||
options:
|
||
a: "/info"
|
||
b: "/status"
|
||
c: "/stats"
|
||
d: "/session"
|
||
correct: "b"
|
||
explanation: |
|
||
The `/status` command shows session information.
|
||
|
||
Output includes:
|
||
- Model being used
|
||
- Context usage percentage
|
||
- Session cost
|
||
- Token counts
|
||
|
||
Example output:
|
||
`Model: Sonnet | Ctx: 45.2k | Cost: $1.23 | Ctx(u): 42.0%`
|
||
|
||
Use `/stats` for usage statistics with activity graphs.
|
||
doc_reference:
|
||
file: "guide/ultimate-guide.md"
|
||
section: "10.1 Commands Table"
|
||
anchor: "#built-in-commands"
|
||
|
||
- id: "10-011"
|
||
difficulty: "senior"
|
||
profiles: ["senior", "power"]
|
||
question: "What flag limits maximum API spend in headless mode?"
|
||
options:
|
||
a: "--cost-limit"
|
||
b: "--max-budget-usd"
|
||
c: "--spending-cap"
|
||
d: "--budget"
|
||
correct: "b"
|
||
explanation: |
|
||
The `--max-budget-usd` flag sets maximum API spend (only with `--print`):
|
||
|
||
```bash
|
||
claude -p "analyze" --max-budget-usd 5.00
|
||
```
|
||
|
||
This prevents runaway costs in automated pipelines.
|
||
The operation stops if the budget is exceeded.
|
||
doc_reference:
|
||
file: "guide/ultimate-guide.md"
|
||
section: "10.3 Configuration Reference"
|
||
anchor: "#cli-flags-reference"
|
||
|
||
- id: "10-012"
|
||
difficulty: "junior"
|
||
profiles: ["junior", "senior", "power"]
|
||
question: "How do you run a shell command directly from Claude Code prompt?"
|
||
options:
|
||
a: "shell: command"
|
||
b: "!command"
|
||
c: "$command"
|
||
d: "run command"
|
||
correct: "b"
|
||
explanation: |
|
||
The `!command` prefix runs a shell command directly.
|
||
|
||
Quick actions:
|
||
- `!command` - Run shell command
|
||
- `@filename` - Reference file
|
||
- `Ctrl+C` - Cancel operation
|
||
- `Ctrl+R` - Retry last
|
||
|
||
Example: `!git status` runs git status without Claude interpreting it.
|
||
doc_reference:
|
||
file: "guide/ultimate-guide.md"
|
||
section: "10.1 Commands Table"
|
||
anchor: "#quick-actions"
|
||
|
||
- id: "10-013"
|
||
difficulty: "senior"
|
||
profiles: ["senior", "power"]
|
||
question: "What is the correct way to resume a specific session by ID?"
|
||
options:
|
||
a: "claude --session abc123"
|
||
b: "claude -r abc123"
|
||
c: "claude --load abc123"
|
||
d: "claude -s abc123"
|
||
correct: "b"
|
||
explanation: |
|
||
Use `-r` or `--resume` to resume a specific session:
|
||
|
||
```bash
|
||
claude -r abc123 # Resume session abc123
|
||
claude --resume abc123 # Same as above
|
||
claude -c # Resume last session (short)
|
||
claude --continue # Resume last session (long)
|
||
```
|
||
|
||
Combine with `-p` for scripting: `claude -r abc123 -p "check status"`
|
||
doc_reference:
|
||
file: "guide/ultimate-guide.md"
|
||
section: "10.3 Configuration Reference"
|
||
anchor: "#cli-flags-reference"
|
||
|
||
- id: "10-014"
|
||
difficulty: "power"
|
||
profiles: ["power"]
|
||
question: "What is the recommended .gitignore pattern for Claude Code files?"
|
||
options:
|
||
a: "Ignore all .claude/ contents"
|
||
b: "Ignore settings.local.json and .claude/CLAUDE.md, keep agents/commands/hooks"
|
||
c: "Don't ignore anything"
|
||
d: "Ignore only agents/"
|
||
correct: "b"
|
||
explanation: |
|
||
Recommended .gitignore:
|
||
|
||
```gitignore
|
||
# Claude Code - Personal (ignore)
|
||
.claude/settings.local.json
|
||
.claude/CLAUDE.md
|
||
.claude/.serena/
|
||
|
||
# Claude Code - Team (keep/commit)
|
||
# .claude/agents/
|
||
# .claude/commands/
|
||
# .claude/hooks/
|
||
# .claude/settings.json
|
||
```
|
||
|
||
This keeps team workflows shared while personal settings stay private.
|
||
doc_reference:
|
||
file: "guide/ultimate-guide.md"
|
||
section: "9.11 Common Pitfalls & Best Practices"
|
||
anchor: "#recommended-gitignore"
|
||
|
||
- id: "10-015"
|
||
difficulty: "junior"
|
||
profiles: ["junior", "senior", "power"]
|
||
question: "What is the daily workflow morning setup according to the guide?"
|
||
options:
|
||
a: "Just start coding"
|
||
b: "Git pull, /status, load project memory, review yesterday's progress"
|
||
c: "Run all tests first"
|
||
d: "Clear all context"
|
||
correct: "b"
|
||
explanation: |
|
||
Morning setup workflow:
|
||
|
||
1. Git pull latest changes
|
||
2. Review context with `/status`
|
||
3. Load project memory (`/sc:load` if using Serena)
|
||
4. Review yesterday's progress
|
||
|
||
This ensures you start with fresh code and full context awareness.
|
||
doc_reference:
|
||
file: "guide/ultimate-guide.md"
|
||
section: "10.6 Daily Workflow & Checklists"
|
||
anchor: "#daily-workflow-pattern"
|
||
|
||
- id: "10-016"
|
||
difficulty: "senior"
|
||
profiles: ["senior", "power"]
|
||
question: "What flag allows Claude's tools to access directories outside the current working directory?"
|
||
options:
|
||
a: "--include-dir"
|
||
b: "--add-dir"
|
||
c: "--context-dir"
|
||
d: "--load-dir"
|
||
correct: "b"
|
||
explanation: |
|
||
Use `--add-dir` to allow tool access to additional directories:
|
||
|
||
```bash
|
||
claude --add-dir ../shared ../utils
|
||
claude --add-dir packages/api
|
||
```
|
||
|
||
By default, Claude can only access files in the current working directory.
|
||
Use --add-dir to extend permissions to other directories (e.g., shared libraries in a monorepo).
|
||
doc_reference:
|
||
file: "guide/ultimate-guide.md"
|
||
section: "10.3 Configuration Reference"
|
||
anchor: "#cli-flags-reference"
|
||
|
||
- id: "10-017"
|
||
difficulty: "power"
|
||
profiles: ["power"]
|
||
question: "When Sequential Thinking MCP seems slow or unresponsive, what should you expect?"
|
||
options:
|
||
a: "It's broken and needs restart"
|
||
b: "10-30 second responses are normal due to significant compute"
|
||
c: "Switch to a different MCP"
|
||
d: "Reduce the query complexity"
|
||
correct: "b"
|
||
explanation: |
|
||
Sequential Thinking uses significant compute - expect 10-30 second responses.
|
||
|
||
This is not an error, just be patient.
|
||
|
||
Tips for Sequential:
|
||
- Works best with specific, well-defined problems
|
||
- Good: "Debug why user authentication fails on mobile"
|
||
- Bad: "Make the app better"
|
||
|
||
The longer response time reflects deeper analysis.
|
||
doc_reference:
|
||
file: "guide/ultimate-guide.md"
|
||
section: "10.4 Troubleshooting"
|
||
anchor: "#sequential-thinking-mcp-issues"
|
||
|
||
- id: "10-018"
|
||
difficulty: "junior"
|
||
profiles: ["junior", "senior", "power"]
|
||
question: "What shortcut opens an external editor for composing long text input?"
|
||
options:
|
||
a: "Ctrl+E"
|
||
b: "Ctrl+G"
|
||
c: "Ctrl+O"
|
||
d: "Ctrl+L"
|
||
correct: "b"
|
||
explanation: |
|
||
`Ctrl+G` opens an external editor for composing long text.
|
||
|
||
Useful input shortcuts:
|
||
- `Ctrl+A`: Jump to beginning of line
|
||
- `Ctrl+E`: Jump to end of line
|
||
- `Ctrl+W`: Delete previous word
|
||
- `Ctrl+G`: Open external editor
|
||
- `Tab`: Autocomplete file paths
|
||
|
||
The external editor allows comfortable editing of complex prompts.
|
||
doc_reference:
|
||
file: "guide/ultimate-guide.md"
|
||
section: "10.2 Keyboard Shortcuts"
|
||
anchor: "#input--navigation"
|