AI Prompt Commands & Parameters
AI tools are increasingly driven by what you type, not what you click: slash commands steer coding agents, @ mentions attach context, and generation parameters shape images and video. This section is a structured reference for that typed layer - the commands, parameters, and prompt syntax of AI tools - built the same way as our keyboard shortcuts reference: every entry is a literal token taken from official documentation, with a description and a usage note, and every table links its source.
This is deliberately not a prompt gallery. There are no prompt templates or example prompts here; those age fast and are easy to find elsewhere. What is hard to find is a maintained, machine-readable list of what a tool actually accepts: which slash commands exist in Claude Code today, what each documented Midjourney parameter does, or how A1111 attention weighting is spelled. That is what these pages hold, and the same data is available as JSON through our public API.
Keyboard shortcuts for these tools are kept in the Shortcuts section.
Platforms
OpenAI-Compatible Endpoints
Tools in this reference that serve or accept the OpenAI API shape - point an OpenAI SDK at them by swapping the base URL. Details and every parameter are on each platform page.
| Platform | What is compatible | Documented tokens on our page |
|---|---|---|
| DeepSeek | Chat completions; beta base URL adds prefix and FIM completion | prefix, suffix, deepseek-reasoner, response_format |
| Qwen (Alibaba Model Studio) | OpenAI-compatible chat endpoint that also accepts DeepSeek, Kimi, GLM and MiniMax model names | enable_thinking, thinking_budget, search_options (extra_body) |
| LM Studio | Local server exposing /v1/chat/completions, /v1/completions, /v1/embeddings, /v1/responses, /v1/models | http://localhost:1234/v1, temperature, maxTokens, structured |
| Nous Hermes | Hermes 4 served locally speaks the standard chat shape; tool calls use Hermes tag protocol | thinking=True, <tool_call>, <tools> |
The comparisons Claude Code vs Cursor and Claude Code vs Codex CLI pair the same task across tools.
Compact References
Smaller command sets live directly on this page.
Suno AI · Source: Suno — How to Make a Song with Suno (suno.com Hub guide, updated 2026-08-12)
Song structure tags (4)
| Command / Syntax | What it does | Notes |
|---|---|---|
| [Intro] | Intro section | Labels the opening section of lyrics written in Advanced mode; the guide's example structure starts with an intro before the first verse. |
| [Verse] | Verse section | Labels a verse of lyrics you write in Advanced mode; the guide numbers repeats, as in [Verse 1] and a later Verse 2. |
| [Chorus] | Chorus section | Labels the chorus lines; the guide also describes their sound in the style text, such as backing harmonies during chorus sections. |
| [Bridge] | Bridge section | Labels the bridge; the guide's sample structure places it after the second verse, leading into the final chorus. |
Agent Instruction Files · 10 sources
Project instruction files (7)
| Command / Syntax | What it does | Notes |
|---|---|---|
| AGENTS.md | Cross-tool agent instructions | Plain Markdown with no required fields; a monorepo can add another copy in each package, and agents follow the one nearest the edited file. |
| CLAUDE.md | Claude Code project instructions | Goes in the project root or .claude/; while one exists on the path, Claude Code skips AGENTS.md unless the Project instructions setting allows both. |
| CLAUDE.local.md | Personal project instructions | Uncommitted notes for one project, listed in .gitignore; it also counts as a CLAUDE.md, so Claude then skips AGENTS.md by default. |
| AGENTS.override.md | Codex override file | Codex reads it instead of AGENTS.md in the same folder, or in ~/.codex for a temporary global change; remove it to restore the normal file. |
| GEMINI.md | Gemini CLI context file | Copies in the workspace and its parent folders are concatenated and sent with every prompt; AGENTS.md is read only when context.fileName names it. |
| .github/copilot-instructions.md | Copilot repository-wide instructions | Added to every Copilot request made in that repository; Copilot Chat lists the file under References when it was used. |
| CONVENTIONS.md | Aider coding conventions file | Not loaded on its own: add it with /read or aider --read so it stays read-only and can be prompt-cached. |
Rule folders and scoped files (6)
| Command / Syntax | What it does | Notes |
|---|---|---|
| .claude/rules/ | Claude Code modular rules | One topic per Markdown file, found recursively; a paths field in the frontmatter loads a rule only when Claude touches matching files. |
| .cursor/rules | Cursor project rules | Each rule must be an .mdc file, since plain .md files there are ignored; alwaysApply, globs and description decide when Agent includes it. |
| .cursorrules | Cursor legacy rules file | Single root file due to be deprecated; to migrate, paste its text into a new rule set to Always Apply, then delete it. |
| NAME.instructions.md | Copilot path-specific instructions | Stored in or below .github/instructions with an applyTo glob in the frontmatter; on GitHub.com only the cloud agent and code review use them. |
| .devin/rules/*.md | Windsurf (Devin Desktop) rules | Preferred over the older .windsurf/rules folder; each file sets a trigger such as always_on or glob and is capped at 12,000 characters. |
| .clinerules/ | Cline workspace rules | Every Markdown file in it is merged into one rule set and can be toggled on or off; .cline/rules/ is searched as well. |
User-level files (6)
| Command / Syntax | What it does | Notes |
|---|---|---|
| ~/.claude/CLAUDE.md | User-level Claude Code instructions | Applies in all your projects and keeps loading next to a repository AGENTS.md, unlike a project CLAUDE.md. |
| ~/.cursor/rules | Cursor local user rule files | Personal rules for this machine only; user rules typed into Customize sync with your Cursor account instead. |
| ~/.codex/AGENTS.md | Codex global instructions | Read at the start of every Codex run, before the project files from the repository root down; CODEX_HOME moves it. |
| ~/.gemini/GEMINI.md | Gemini CLI global context | Loaded first for all projects; after editing any context file mid-session, /memory reload re-reads them. |
| ~/.codeium/windsurf/memories/global_rules.md | Windsurf global rules | One always-on file shared by every workspace, capped at 6,000 characters. |
| ~/Documents/Cline/Rules | Cline global rules folder | Default on macOS and Linux (Documents\Cline\Rules on Windows); workspace rules win when the two disagree. |
Compatibility patterns (5)
| Command / Syntax | What it does | Notes |
|---|---|---|
| ln -s AGENTS.md CLAUDE.md | Symlink CLAUDE.md to AGENTS.md | Optional since Claude Code v2.1.277, which reads AGENTS.md itself when no CLAUDE.md exists; Windows clones may get the link as plain text. |
| @AGENTS.md | Import AGENTS.md into CLAUDE.md | Place it at the top of a CLAUDE.md and add Claude-only notes below; works where symlinks do not, such as Windows checkouts. |
| context.fileName | Gemini CLI context file names | A settings.json key taking one name or a list; set it to AGENTS.md, with or without GEMINI.md, so Gemini CLI loads that file. |
| read: AGENTS.md | Load AGENTS.md in Aider | A line for .aider.conf.yml, so Aider adds the file at every start without a /read command. |
| project_doc_fallback_filenames | Codex extra instruction file names | Set in ~/.codex/config.toml; names listed there, such as TEAM_GUIDE.md, are tried after AGENTS.override.md and AGENTS.md. |
Frequently Asked Questions
What counts as a prompt command on this site?
A literal, typeable token from official documentation: slash commands, @ mentions, chat variables, generation parameters, and prompt syntax operators. We do not collect prompt templates or example prompts.
How is this different from the keyboard shortcuts pages?
The Shortcuts section covers keys you press and terminal CLI invocations. This section covers what you type inside an AI tool's prompt to control it - commands, parameters, and syntax.
Where does the data come from?
Each platform lists its source: official docs pages, vendor help centers, or the project's own wiki. Every table links its sources, and each row was matched to a line in a linked source.
Developers: this data is available as JSON at /api/v1/prompts/all.json and per platform at /api/v1/prompts/<key>.json.