Interactive Mode Guide
When run without required arguments, many prompts-gpt commands enter an interactive mode that guides you through the options with prompts and selections.
How It Works
Interactive mode activates when:
- No prompt file is specified for
runorsweep - No provider is specified and multiple are available
- The
quickstartcommand is run
Interactive Run
prompts-gpt run
The CLI will prompt you:
? Select a prompt file:
❯ .prompts-gpt/code-review.md
.prompts-gpt/security-review.md
.prompts-gpt/performance-check.md
? Select a provider:
❯ codex (v1.0.12)
claude (v1.0.30)
cursor (v0.46.0)
? Select a model (or press Enter for default):
❯ gpt-5.5
gpt-5.4
gpt-5.3
Interactive Sweep
prompts-gpt sweep
Additional prompts:
? Select a sweep file:
❯ .prompts-gpt/sweeps/security-audit.md
.prompts-gpt/sweeps/performance-audit.md
? Number of iterations: 5
? Enable self-evaluation? Yes
Interactive Quickstart
prompts-gpt quickstart
Walks through:
- Token setup (prompts for token input)
- Provider detection
- Prompt selection or generation
- First run execution
Interactive Setup
prompts-gpt setup
Configures .prompts-gpt/config.json:
? Default agent profile:
❯ router (auto-detect)
codex
claude
cursor
? Provider priority order: codex, claude, cursor
? Default timeout (seconds): 900
? Agent targets to sync:
✓ codex
✓ cursor
✓ claude-code
copilot
continue
Interactive Generate
prompts-gpt generate
Prompts for generation parameters:
? What should the prompt accomplish?
> Review authentication middleware for security issues
? Additional context (optional):
> Next.js App Router, Prisma ORM, Supabase Auth
? Target tool:
❯ Codex
Claude
Cursor
Copilot
? Include web search for best practices? Yes
Disabling Interactive Mode
In CI or scripts, interactive mode is auto-disabled when:
CIenvironment variable is set--jsonflag is used- stdin is not a TTY
You can also explicitly provide all required arguments:
prompts-gpt run .prompts-gpt/review.md --provider codex --model gpt-5.5