AI Agent Workflow (Workflow & Productivity)

# Continue the last conversation
claude --continue

# Resume a specific session
claude --resume <session-name>

# Name the session during the conversation
/rename stripe-integration

# ~/.zshrc or ~/.bashrc
alias c='claude'
alias cc='claude --continue'
alias cr='claude --resume'
alias g='gemini'
alias cx='codex'

"Analyze the changes, write an appropriate commit message, then commit"

"Create a draft PR from the current branch's changes.
Make the title summarize the changes, and list the key changes in the body."

# Work on multiple branches simultaneously
git worktree add ../myapp-feature-auth feature/auth
git worktree add ../myapp-hotfix hotfix/critical-bug

# Independent AI sessions per worktree
Tab 1: ~/myapp-feature-auth → new feature development
Tab 2: ~/myapp-hotfix → urgent bug fix
Tab 3: ~/myapp (main) → keep main branch

1. "Run gh pr checkout 123 and summarize this PR's changes"
2. "Analyze changes in src/auth/middleware.ts. Check for security issues or performance problems"
3. "Is there a way to make this logic more efficient?"
4. "Apply the improvements you suggested and run tests"

# Gemini: large-scale analysis
> ask-gemini "@src/ Analyze the structure of the entire codebase"

# Codex: run commands
> shell "docker-compose up -d"
> shell "npm test && npm run build"

# Disable unused MCP servers
/mcp

# Recommended numbers
# - MCP servers: fewer than 10
# - Active tools: fewer than 80

[Claude] Plan → [Gemini] Analysis/research → [Claude] Write code → [Codex] Run/test → [Claude] Synthesize results

1. [Claude] Design API spec using the skill
2. [Gemini] ask-gemini "@src/ Analyze existing API patterns" - large-scale codebase analysis
3. [Claude] Implement code based on the analysis
4. [Codex] shell "npm test && npm run build" - test and build
5. [Claude] Create final report

"Work using TDD. First write a failing test,
then write code that makes the test pass."

# The AI:
# 1. Write a failing test
# 2. git commit -m "Add failing test for user auth"
# 3. Write minimal code to pass the test
# 4. Run tests → confirm they pass
# 5. git commit -m "Implement user auth to pass test"

FROM ubuntu:22.04
RUN apt-get update && apt-get install -y \
    curl git tmux vim nodejs npm python3 python3-pip
RUN curl -fsSL https://claude.ai/install.sh | sh
WORKDIR /workspace
CMD ["/bin/bash"]

# Build and run the container
docker build -t ai-sandbox .
docker run -it --rm \
  -v $(pwd):/workspace \
  -e ANTHROPIC_API_KEY=$ANTHROPIC_API_KEY \
  ai-sandbox

# Do experimental work inside the container

/context  # Check usage
/clear    # Reset context

# Or create HANDOFF.md and start a new session

Esc Esc  # Cancel the last task immediately

# Check MCP/tool counts
/mcp

# Disable unnecessary MCP servers
# Reset context

=== Essential commands ===
/clear      reset context
/context    check usage
/usage      check tokens
/init       generate project description file
!command    run immediately

=== Shortcuts ===
Esc Esc     cancel task
Ctrl+R      search history
Shift+Tab×2 plan mode
Ctrl+B      background

=== CLI flags ===
--continue  continue conversation
--resume    resume session
-p "prompt" headless mode

=== Multi-Agent ===
Claude      plan/code generation
Gemini      large-scale analysis
Codex       run commands

=== Troubleshooting ===
Context overloaded → /clear
Cancel task → Esc Esc
Performance degradation → check /context