docs: clarify CLI command intent and add output routing

- **Intent Clarification**: Distinguished analysis (read-only) from execution (modifies code) commands
- **Output Routing**: Added comprehensive output destination logic for all CLI commands
- **Scratchpad Integration**: Introduced `.workflow/.scratchpad/` for non-session-specific outputs

## Analysis Commands (Read-Only)
- `/cli:analyze`, `/cli:chat`, `/cli:mode:*`: Added "Core Behavior" emphasizing read-only nature
- Added "MODE: analysis" in command templates
- Examples clearly show recommendations vs. code changes

## Execution Commands (Modifies Code)
- `/cli:execute`: Added ⚠️ warnings about code modification
- `/cli:codex-execute`: Clarified multi-stage execution with detailed output structure
- Both commands now have "Output Routing" sections

## Workflow Architecture Updates
- Added `.workflow/.scratchpad/` directory definition
- Output routing logic: session-relevant → `.chat/`, otherwise → `.scratchpad/`
- File naming pattern: `[command-type]-[brief-description]-[timestamp].md`
- Examples for both analysis and implementation commands

## Key Improvements
- Prevents confusion between analysis and implementation operations
- Solves output loss when no active session exists
- Prevents unrelated analyses from cluttering session history
- Provides centralized location for ad-hoc outputs

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

.claude/commands/cli/chat.md

@@ -14,10 +14,18 @@ allowed-tools: SlashCommand(*), Bash(*)
 
 ## Purpose
 
-Direct Q&A interaction with CLI tools for codebase analysis.
+Direct Q&A interaction with CLI tools for codebase analysis. **Analysis only - does NOT modify code**.
 
+**Intent**: Ask questions, get explanations, understand codebase structure
 **Supported Tools**: codex, gemini (default), qwen
 
+## Core Behavior
+
+1. **Conversational Analysis**: Direct question-answer interaction about codebase
+2. **Read-Only**: This command ONLY provides information and analysis
+3. **No Code Modification**: Results are explanations and insights
+4. **Flexible Context**: Choose specific files or entire codebase
+
 ## Parameters
 
 - `<inquiry>` (Required) - Question or analysis request
@@ -31,8 +39,9 @@ Direct Q&A interaction with CLI tools for codebase analysis.
 1. Parse tool selection (default: gemini)
 2. If `--enhance`: Execute `/enhance-prompt` to expand user intent
 3. Assemble context: `@{CLAUDE.md}` + user-specified files or `--all-files`
-4. Execute CLI tool with assembled context
-5. Optionally save to workflow session
+4. Execute CLI tool with assembled context (read-only, analysis mode)
+5. Return explanations and insights (NO code changes)
+6. Optionally save to workflow session
 
 ## Context Assembly
 
@@ -44,18 +53,69 @@ Direct Q&A interaction with CLI tools for codebase analysis.
 
 For targeted analysis, use `rg` or MCP tools to discover relevant files first, then build precise CONTEXT field.
 
-## Examples
+## Command Template
 
 ```bash
-/cli:chat "analyze the authentication flow"              # Gemini (default)
-/cli:chat --tool qwen "optimize React component"         # Qwen
-/cli:chat --tool codex "review security vulnerabilities" # Codex
-/cli:chat --enhance "fix the login"                      # Enhanced prompt first
-/cli:chat --all-files "find all API endpoints"           # Full codebase context
+cd . && ~/.claude/scripts/gemini-wrapper -p "
+INQUIRY: [user question]
+CONTEXT: @{CLAUDE.md,**/*CLAUDE.md} [inferred or --all-files]
+MODE: analysis
+RESPONSE: Direct answer, explanation, insights (NO code modification)
+"
 ```
 
+## Examples
+
+**Basic Question**:
+```bash
+/cli:chat "analyze the authentication flow"
+# Executes: Gemini analysis
+# Returns: Explanation of auth flow, components involved, data flow
+```
+
+**Architecture Question**:
+```bash
+/cli:chat --tool qwen "how does React component optimization work here"
+# Executes: Qwen architecture analysis
+# Returns: Component structure explanation, optimization patterns used
+```
+
+**Security Analysis**:
+```bash
+/cli:chat --tool codex "review security vulnerabilities"
+# Executes: Codex security analysis
+# Returns: Vulnerability assessment, security recommendations (NO automatic fixes)
+```
+
+**Enhanced Inquiry**:
+```bash
+/cli:chat --enhance "explain the login issue"
+# Step 1: Enhance to expand login context
+# Step 2: Analysis with expanded understanding
+# Returns: Detailed explanation of login flow and potential issues
+```
+
+**Broad Context**:
+```bash
+/cli:chat --all-files "find all API endpoints"
+# Executes: Analysis across entire codebase
+# Returns: List and explanation of API endpoints (NO code generation)
+```
+
+## Output Routing
+
+**Output Destination Logic**:
+- **Active session exists AND query is session-relevant**:
+  - Save to `.workflow/WFS-[id]/.chat/chat-[timestamp].md`
+- **No active session OR unrelated query**:
+  - Save to `.workflow/.scratchpad/chat-[description]-[timestamp].md`
+
+**Examples**:
+- During active session `WFS-api-refactor`, asking about API structure → `.chat/chat-20250105-143022.md`
+- No session, asking about build process → `.scratchpad/chat-build-process-20250105-143045.md`
+
 ## Notes
 
 - Command templates and file patterns: see intelligent-tools-strategy.md (loaded in memory)
-- Active workflow session: results saved to `.workflow/WFS-[id]/.chat/`
-- No session: results returned directly
+- Scratchpad directory details: see workflow-architecture.md
+- Scratchpad conversations preserved for future reference