Prompter

Turn rough notes into structured prompts from the command line:

echo "write a bash backup script" | prompter refine

Prompter refines and critiques rough prompts, restructures Markdown, applies reusable catalog prompts, and builds image prompts offline. Generated text goes to stdout; errors, progress, and timing go to stderr.

Install

# Homebrew
brew install dotcommander/tap/prompter

# Go
go install github.com/dotcommander/prompter@latest

Gemini works with Google Application Default Credentials. Other remote providers use their standard API-key environment variables, such as OPENAI_API_KEY or GROQ_API_KEY; see the providers guide for per-provider details, including how to use GEMINI_API_KEY with the AI Studio endpoint.

Commands

Command	Purpose
prompter refine [input]	Improve rough prompt text.
prompter critique [input]	Analyze ambiguity and missing constraints without rewriting.
prompter rewrite [input]	Restructure Markdown or documentation.
prompter apply <name> [input]	Apply an exact catalog prompt name or alias.
prompter browse	Search the local prompt vault interactively.
prompter image <subject>	Build an image-generation prompt offline.
prompter configure	Configure Prompter, or print resolved settings non-interactively.
prompter models refresh	Cache a short list of affordable model choices per provider from Models.dev, OpenRouter, and the local OMLX server.
prompter prompts status	Classify installed starter prompts as current, upgradeable, customized, staged, or missing.
prompter prompts upgrade	Install missing prompts and stage versioned replacements without overwriting existing files.

Bare prompter displays help on an interactive terminal. Piped input defaults to prompter refine when no command is given.

prompter refine -s code "design a retry loop"
prompter critique "summarize this better"
prompter rewrite --file notes.md --mode clean
prompter prompts upgrade --dry-run
cat spec.md | prompter apply system-architect > architecture.md
prompter image "desert observatory" --count 3
prompter browse
prompter configure
prompter models refresh

Use prompter <command> --help for command flags. Common LLM overrides include -p/--provider, -m/--model, -f/--file, -o/--output, -c/--copy, --stream, --dry-run, --base-url, and -v/--verbose.

Behavior

• refine, critique, rewrite, and apply call the configured LLM provider.
• image is deterministic and offline; it builds prompt text but does not generate an image.
• browse copies the selected prompt to the clipboard and writes it to stdout. On first run it seeds an empty vault with starter prompts, any of which can be used via apply <name-or-alias>.
• apply strips YAML frontmatter. Prompts may declare output validation; validated calls buffer output and reject invalid responses.
• Streaming can emit partial text before a provider reports truncation. Always discard streamed output when Prompter exits nonzero.

Configuration precedence is:

CLI flags > environment variables > ~/.config/prompter/config.json > defaults

Set PROMPTER_DEFAULT_COPY=true or "default_copy": true to copy buffered results automatically.

Documentation

• Setup
• Flags
• Common tasks
• Prompt files
• Providers
• Troubleshooting
• Contributor architecture