{"slug": "the-grammar-of-coding-agents", "title": "The Grammar of Coding Agents", "summary": "OpenAI Codex, Claude Code, OpenCode, Cursor, and GitHub Copilot increasingly share building blocks like slash commands, skills, and subagents but expose them through different command languages, according to a comparative analysis verified June 19, 2026. The analysis details how each product handles persistent objectives, planning, parallel subagents, and code review, with Claude Code offering the broadest built-in command catalog and Codex using a distinct $ prefix for skills.", "body_md": "# The Grammar of Coding Agents\n\n**OpenAI Codex, Claude Code, OpenCode, Cursor, and GitHub Copilot**\n\n**Snapshot verified:** June 19, 2026**Scope:** Interactive command languages, context sigils, skills, subagents, planning, review, insights, and differences between terminal, desktop, IDE, and cloud harnesses.\n\nThis is a living reference. Command availability can vary by version, operating system, plan, experimental flags, and connected extensions. In every harness, typing\n\n`/`\n\nin the active composer is the best final check for what is available in that exact installation.\n\n## Executive summary\n\nThe five product families increasingly share the same building blocks - slash commands, reusable skills, project instructions, MCP, custom agents, shell execution, and context management - but they expose them through different command languages.\n\n**Codex** has the clearest use of`$`\n\n:`$skill-name`\n\nexplicitly invokes a skill, while`$app-slug`\n\nattaches an app/connector.`@`\n\nis primarily workspace/file context, though recent Codex releases are moving toward a unified mentions menu. Codex has`/goal`\n\n, but parallel subagents are requested in natural language rather than through`/fleet`\n\n.**Claude Code** has the broadest built-in command catalog and the richest collection of packaged workflows. Skills are invoked as`/skill-name`\n\n, not with`$`\n\n. Its stand-out commands include`/insights`\n\n,`/batch`\n\n,`/code-review`\n\n,`/loop`\n\n,`/ultraplan`\n\n, and`/workflows`\n\n.**OpenCode** has a smaller built-in slash-command surface but a very flexible customization model.`@`\n\ncan reference both files and agents; custom commands support`$ARGUMENTS`\n\n,`$1`\n\n,`$2`\n\n, shell-output injection, file references, agent selection, and forced subtask execution.**Cursor combines an IDE, terminal CLI, and Cloud Agents around one rules, skills, MCP, and subagent system.** Skills and custom subagents use`/name`\n\n;`@`\n\nattaches files, folders, docs, terminals, prior chats, diffs, browser state, and manual rules. Cursor CLI uses`/shell`\n\nrather than a documented`!`\n\nshortcut, and`& prompt`\n\nhands a local CLI task to a Cloud Agent.**GitHub Copilot must be split by harness.** Copilot CLI and the Copilot app use a modern agent runtime with`/fleet`\n\n,`/chronicle`\n\n,`/research`\n\n, skills, plugins, and custom agents. VS Code Chat uses a different interaction model centered on`/explain`\n\n,`/fix`\n\n,`#file`\n\n, and`@workspace`\n\n. GitHub.com Chat has another, smaller set of conversation commands and`@`\n\nattachments.\n\n### How to read the comparison\n\nThe closest equivalents are:\n\n| Intent | Codex | Claude Code | OpenCode | Cursor | GitHub Copilot CLI |\n|---|---|---|---|---|---|\n| Persistent objective | `/goal` |\nNatural language, plan/workflow; no direct equivalent | Natural language; no direct equivalent | Plan/checkpoints; no persistent `/goal` object |\nPlan/autopilot; no direct `/goal` equivalent |\n| Plan before editing | `/plan` |\n`/plan` or `/ultraplan` |\nSwitch to Plan agent |\nPlan mode: `Shift+Tab` , `/plan` , or `--mode=plan` |\n`/plan` ; `Shift+Tab` can select plan mode |\n| Parallel subagents | Ask explicitly: “spawn agents in parallel” | `/batch` , subagents, agent teams, workflows |\n`@subagent` , automatic delegation, or `subtask: true` |\nAutomatic or `/subagent-name` ; ask for parallel work; `/best-of-n` uses worktrees |\n`/fleet` |\n| Inspect agent workers | `/agent` |\n`/agents` , `/tasks` , `claude agents` |\nChild-session navigation | Agent UI; background output under `~/.cursor/subagents/` ; Cloud Agents dashboard |\n`/tasks` , `/agents` , `/subagents` |\n| Session insights | `/usage` , `/status` ; no `/insights` equivalent |\n`/insights` , `/usage` , `/team-onboarding` |\nNo built-in equivalent | `/context` ; team analytics, but no personal `/insights` equivalent |\n`/chronicle` , `/usage` |\n| Code review | `/review` , `/diff` |\n`/code-review` , `/review` , `/security-review` , `/simplify` |\nPlan/review agent or custom command | CLI `Ctrl+R` ; Agent `/review` or `/review-bugbot` ; Bugbot PR review |\n`/review` , `/security-review` , `/diff` , `/pr` |\n| Explicit skill | `$skill-name` or `/skills` picker |\n`/skill-name` |\nAgent loads through the `skill` tool |\n`/skill-name` |\n`/skill-name` ; manage with `/skills` |\n| File context | `@path` , `/mention` |\n`@path` |\n`@path` |\n`@file` , `@folder` , `@Docs` , `@Terminals` , `@Past Chats` , diffs, browser |\nCLI: `@path` ; VS Code: `#file` ; GitHub.com: `@` attachment |\n| Direct shell | `!command` |\n`!command` |\n`!command` |\nCLI `/shell command` , `/sh` , or `/run` ; no documented `!` shortcut |\n`!command` ; `!` alone enters shell mode |\n\n## 1. The sigils: what `/`\n\n, `@`\n\n, `$`\n\n, `#`\n\n, and `!`\n\nmean\n\n| Sigil | Codex | Claude Code | OpenCode | Cursor | GitHub Copilot |\n|---|---|---|---|---|---|\n`/` |\nBuilt-in session commands; enabled skills also appear in the list | Built-in commands, bundled skills, user skills, plugin skills, MCP prompts | Built-in and user-defined custom commands | CLI commands, Agent Skills, custom subagents, and product workflows such as `/worktree` and `/review` |\nCLI: full runtime commands and skills. VS Code/GitHub.com: surface-specific chat commands |\n`@` |\nStable docs: fuzzy workspace/file mention. Recent releases also expose a broader unified mention menu | File mention in normal sessions; in agent view it can target a custom agent or sibling repository | File/reference mention and subagent mention | Files/folders, docs, terminals, past chats, Git diffs, browser context, and manually applied rules | CLI: file context. VS Code: chat participants such as `@workspace` . GitHub.com: attach repos, issues, PRs, files, and extensions. Fleet prompts can name `@custom-agent` |\n`$` |\nFirst-class interactive syntax: `$skill-name` and `$app-slug` |\nNot the normal invocation sigil. `$ARGUMENTS` and positional variables are used inside skill/command definitions |\nNot the normal invocation sigil. `$ARGUMENTS` , `$1` , `$2` , and so on are custom-command template variables |\nNo primary interactive `$` convention; ordinary shell environment variables still use `$VAR` |\nNo primary interactive `$` convention. Skills use `/skill-name` ; shell/environment configuration may still contain ordinary `$VAR` syntax |\n`#` |\nNo primary interactive context convention | In agent view, `#123` can identify an existing PR session |\nNo primary interactive context convention | No primary interactive context convention documented | CLI: `#123` attaches a GitHub issue or PR. VS Code: `#file` , `#selection` , `#project` , and other chat variables |\n`!` |\nRun a local shell command and add the result to context, subject to sandbox/approval rules | Enter shell mode for one command; output enters context | Run a shell command; output enters context | No documented `!` shortcut; Cursor CLI uses `/shell` , `/sh` , or `/run` |\nRun a local shell command bypassing Copilot. `!` alone enters multi-command shell mode |\n\n### Sigil portability at a glance\n\n### Practical consequence\n\nThe most dangerous portability mistake is assuming the same sigil means the same thing:\n\n`@workspace`\n\nis a Copilot VS Code participant, not a portable way to request repository context elsewhere.`$frontend-design`\n\nis a Codex skill invocation, but Claude Code and Copilot normally expect`/frontend-design`\n\n.`$ARGUMENTS`\n\nin Claude Code and OpenCode belongs inside the reusable command definition; it is not how an end user invokes the command.`#file`\n\nis Copilot vocabulary. Codex, Claude Code, and OpenCode normally use`@file`\n\n.- Cursor uses\n`/skill-name`\n\nand`/subagent-name`\n\n; its`@`\n\nmenu is for context and manual rules, not the primary custom-agent selector.\n\n## 2. Capability comparison\n\nLegend: **Yes** = directly documented; **Partial** = available through another mechanism or only some surfaces; **No** = no close built-in equivalent documented.\n\n| Capability | Codex | Claude Code | OpenCode | Cursor | Copilot CLI | Copilot VS Code Chat |\n|---|---|---|---|---|---|---|\n| Rich built-in slash-command catalog | Yes | Yes - largest catalog | Partial - compact built-in set | Yes in CLI; editor also exposes skills and workflow commands | Yes | Partial - task shortcuts |\n| User-defined slash commands | Legacy custom prompts are deprecated in favor of skills | Yes; custom commands merged into skills | Yes; highly configurable command templates | Skills replace older commands; `/migrate-to-skills` converts them |\nYes through skills/plugins | Yes through `.prompt.md` prompt files |\nAgent Skills (`SKILL.md` ) |\nYes | Yes | Yes | Yes; reads Cursor, Agent, Claude, and Codex skill directories | Yes | Yes in agent mode |\n| Explicit skill syntax | `$skill-name` |\n`/skill-name` |\nNo dedicated user sigil documented; agent uses skill tool | `/skill-name` |\n`/skill-name` |\nPrompt files use `/name` ; Agent Skills can be selected/invoked by the agent |\n| Implicit skill selection | Yes | Yes | Yes | Yes unless `disable-model-invocation: true` |\nYes | Yes in agent mode |\n| Project instructions | `AGENTS.md` |\n`CLAUDE.md` |\n`AGENTS.md` |\n`.cursor/rules/*.mdc` , `AGENTS.md` ; CLI also reads root `CLAUDE.md` |\n`.github/copilot-instructions.md` , `AGENTS.md` , and related instruction files |\nSame Copilot instruction system |\n| Custom agents | Yes | Yes | Yes | Yes: `.cursor/agents` , with Claude/Codex-compatible locations |\nYes | Yes |\n| Named agent selection | Configure agents; ask in prompt; `/agent` switches spawned threads |\n`/agents` , `--agent` , agent view `@agent` |\nTab for primary agents; `@subagent` |\n`/subagent-name` or natural language; automatic delegation |\n`/agent` , `--agent` , natural language |\nAgent picker/custom agents |\n| Parallel worker command | No dedicated `/fleet` ; ask explicitly |\n`/batch` ; also teams/workflows |\nNo dedicated fleet command; delegation and subtasks | Ask for parallel work; multiple Task calls; `/best-of-n` for parallel worktrees |\n`/fleet` |\nAgent mode supports subagents, but not the CLI `/fleet` vocabulary |\n| Persistent goal object | `/goal` |\nNo direct equivalent | No direct equivalent | No direct equivalent | No direct equivalent | No direct equivalent |\n| Planning mode | `/plan` |\n`/plan` , `/ultraplan` |\nBuilt-in Plan agent | Editor Plan mode; CLI `/plan` , `--plan` , `--mode=plan` |\n`/plan` , plan mode |\nPlan mode |\n| Side question without polluting main history | `/side` or `/btw` |\n`/btw` |\nStart/switch session | Ask mode: `/ask` or `--mode=ask` |\n`/ask` experimental |\nStart a new chat or use another chat surface |\n| Built-in insights from history | No dedicated `/insights` |\n`/insights` |\nNo | `/context` ; team analytics only |\n`/chronicle` |\nSynced agent-session search; no equivalent VS Code slash command |\n| Built-in deep research | Web search/tools; no dedicated slash command in core catalog | `/deep-research` |\nTools/custom agents | Explore subagent and web search tools; no dedicated research slash command documented | `/research` |\n`@github #web` for web-enabled questions |\n| Built-in code review | `/review` |\nMultiple review workflows | Custom/review agent pattern | CLI `Ctrl+R` ; `/review` , `/review-bugbot` , and Bugbot PR review |\n`/review` , `/security-review` |\nReview actions and Copilot code review |\n| Scheduled prompts | Codex app automations | `/loop` , routines/workflows |\nPlugins/external scheduling | No comparable built-in scheduler documented | `/every` , `/after` experimental |\nNot the same interactive mechanism |\n| Session rollback with file restoration | Git-based inspection; no general `/undo` in the app catalog |\n`/rewind` /`/undo` |\n`/undo` and `/redo` require Git |\nEditor checkpoints; CLI `/rewind` |\n`/undo` /`/rewind` , tool-layer tracking |\nCheckpoints in supported agent modes |\n| MCP | Yes | Yes | Yes | Yes across editor, CLI, and Cloud Agents | Yes, GitHub MCP built in | Yes |\n| Plugins | Yes | Yes | Yes | Yes; CLI `/plugin` and marketplaces |\nYes | Extensions/customizations, but not identical to CLI plugins |\n| Local sandbox/permissions | Yes | Yes | Yes | Yes; CLI `/sandbox` , permissions and approval modes |\nYes | Workspace trust and agent tool approvals |\n| Cloud delegation | Codex Cloud | Claude Code on the web | Self-hosted server/web; provider-dependent | Cloud Agents; CLI `& prompt` ; local `/in-cloud` subagent |\n`/delegate` , `--cloud` , cloud agent |\nCan delegate to Copilot cloud agent |\n\n## 3.OpenAI Codex\n\n### Harnesses\n\n| Harness | Best use | Command-language notes |\n|---|---|---|\nCodex CLI |\nTerminal-native local work, scripting, shell-heavy debugging, subagents | Richest Codex slash catalog; `@` , `$` , and `!` are all meaningful |\nCodex app |\nParallel threads, worktrees, automations, Git review, artifacts | Smaller slash list, visual project/session management, shared skills |\nCodex IDE extension |\nEditor context, selections, local/cloud handoff | Surface-specific `/local` , `/cloud` , `/auto-context` ; shared config with CLI |\nCodex Cloud/web |\nBackground work in configured cloud environments | Usually delegated from app, IDE, CLI, or web rather than controlled through the full local slash catalog |\n\n### High-value Codex CLI commands\n\n| Category | Commands |\n|---|---|\n| Project/session | `/init` , `/new` , `/resume` , `/fork` , `/archive` , `/delete` , `/clear` , `/title` |\n| Context | `/compact` , `/status` , `/usage` , `/memories` , `/mention` |\n| Work control | `/plan` , `/goal` , `/side` , `/btw` , `/ps` , `/stop` |\n| Agents/extensions | `/agent` , `/skills` , `/apps` , `/plugins` , `/hooks` , `/mcp` , `/import` |\n| Model/behavior | `/model` , `/fast` , `/personality` , `/permissions` , `/experimental` |\n| Review/output | `/diff` , `/review` , `/copy` , `/raw` , `/theme` , `/statusline` |\n| Account/support | `/feedback` , `/logout` , `/debug-config` , `/exit` , `/quit` |\n\n### Codex app slash commands\n\n`/feedback`\n\n, `/goal`\n\n, `/init`\n\n, `/mcp`\n\n, `/plan`\n\n, `/review`\n\n, `/status`\n\n### Codex IDE extension slash commands\n\n`/auto-context`\n\n, `/cloud`\n\n, `/cloud-environment`\n\n, `/feedback`\n\n, `/goal`\n\n, `/local`\n\n, `/review`\n\n, `/status`\n\n### Skills, apps, and context\n\n- Use\n`$skill-name`\n\nto explicitly invoke a skill. - Use\n`/skills`\n\nto browse available skills; enabled skills also appear in the slash menu. `/apps`\n\ninserts a connector as`$app-slug`\n\n.- Use\n`@path`\n\nor`/mention`\n\nfor file context. - Use\n`!command`\n\nfor a local shell command.\n\n### Subagent behavior\n\nCodex supports parallel subagents in the CLI and app, but the trigger is an explicit natural-language request rather than `/fleet`\n\n. For example:\n\n```\nSpawn three subagents in parallel: one for security, one for test gaps,\nand one for maintainability. Wait for all three and merge the findings.\n```\n\n`/agent`\n\ndoes not itself mean “start a fleet”; it switches between already spawned agent threads.\n\n### Distinctive strengths\n\n- Persistent\n`/goal`\n\nstate with pause/resume/clear behavior. - First-class\n`$`\n\nsyntax for skills and apps. - Tight local/cloud handoff across CLI, app, and IDE.\n- App-centric parallel threads, worktrees, automations, artifact previews, and Git review.\n\n## 4.Claude Code\n\n### Harnesses\n\nClaude Code is available in the terminal, IDE, desktop app, and browser. Anthropic describes the CLI as the most complete surface; desktop and IDE trade some CLI-only capabilities for visual review and tighter editor integration.\n\n### High-value Claude Code commands\n\n| Category | Commands |\n|---|---|\n| Setup/config | `/init` , `/memory` , `/mcp` , `/permissions` , `/agents` , `/skills` , `/config` , `/model` , `/effort` |\n| Planning | `/plan` , `/ultraplan` , `/advisor` |\n| Parallel work | `/agents` , `/tasks` , `/background` , `/batch` , `/fork` , `/workflows` |\n| Context/session | `/context` , `/compact` , `/btw` , `/clear` , `/resume` , `/branch` , `/teleport` , `/remote-control` |\n| Review/verification | `/diff` , `/code-review` , `/review` , `/security-review` , `/simplify` , `/verify` |\n| Research/diagnosis | `/deep-research` , `/debug` , `/doctor` |\n| Automation | `/loop` , `/workflows` , `/autofix-pr` |\n| Insights/usage | `/insights` , `/usage` , `/stats` , `/team-onboarding` |\n\n### Commands that are especially distinctive\n\n`/insights`\n\nanalyzes Claude Code sessions for project areas, interaction patterns, and friction points.`/batch`\n\ndecomposes a large change into roughly 5-30 worktree-isolated subagents, each expected to implement, test, and open a pull request.`/code-review`\n\ncan review, fix, comment on a PR, or run an ultra cloud review.`/simplify`\n\nruns multiple cleanup-oriented review agents.`/loop`\n\nruns a repeated or self-paced task.`/ultraplan`\n\nand`/ultrareview`\n\nconnect terminal work to deeper cloud planning/review.`/team-onboarding`\n\nanalyzes recent usage and generates a teammate onboarding guide.\n\n### Skills and custom commands\n\n- Invoke a skill as\n`/skill-name`\n\n. - User skills live under\n`.claude/skills/<name>/SKILL.md`\n\n. - Older\n`.claude/commands/*.md`\n\nfiles still work, but Anthropic now treats custom commands as skills. `$ARGUMENTS`\n\nand positional parameters are used inside skill definitions.- MCP prompts appear as\n`/mcp__<server>__<prompt>`\n\n.\n\n### Context and shell syntax\n\n`@path`\n\nopens file-path autocomplete and attaches file context.`!command`\n\nruns a command directly and adds its output to the conversation.- In the multi-agent view,\n`@agent-name`\n\ncan choose a custom agent and`@repo`\n\ncan target a sibling repository.\n\n### Subagent choices\n\nClaude Code has several distinct parallelism mechanisms:\n\n**Subagents** for isolated delegated tasks.**Agent view**(`claude agents`\n\n) for multiple background sessions.**Agent teams** for coordinated sessions that communicate.**Dynamic workflows** for background multi-phase orchestration.as a packaged large-change workflow using worktrees and subagents.`/batch`\n\n### Distinctive strengths\n\n- The broadest command vocabulary.\n- Strong history analysis through\n`/insights`\n\n. - Several levels of parallel execution, from one-off subagents to teams and workflows.\n- Rich packaged workflows for review, debugging, research, verification, onboarding, and scheduling.\n\n## 5.OpenCode\n\n### Harnesses\n\nOpenCode is available as a TUI, desktop app, IDE extension, and browser application. It also supports ACP clients. ACP carries most OpenCode capabilities, although some built-in commands such as `/undo`\n\nand `/redo`\n\nare not supported there.\n\n### Built-in TUI commands\n\n| Command | Purpose |\n|---|---|\n`/connect` |\nAdd/configure a model provider |\n`/compact` or `/summarize` |\nCompact the current session |\n`/details` |\nToggle tool-execution detail |\n`/editor` |\nCompose in an external editor |\n`/exit` , `/quit` , `/q` |\nExit |\n`/export` |\nExport the conversation to Markdown |\n`/help` |\nOpen help |\n`/init` |\nCreate or update `AGENTS.md` |\n`/models` |\nSelect/list models |\n`/new` or `/clear` |\nStart a new session |\n`/redo` |\nRestore an undone turn and file changes |\n`/sessions` , `/resume` , `/continue` |\nBrowse and switch sessions |\n`/share` , `/unshare` |\nPublish or remove a shared session |\n`/themes` |\nSelect a theme |\n`/thinking` |\nShow or hide reasoning blocks |\n`/undo` |\nRevert the latest turn and file changes |\n\n### Custom commands\n\nOpenCode's small built-in list is offset by a strong custom-command format:\n\n- Define Markdown commands in\n`.opencode/commands/`\n\n. - Use\n`$ARGUMENTS`\n\n,`$1`\n\n,`$2`\n\n, and so on for parameters. - Inject shell output into a command prompt.\n- Include files with\n`@path`\n\n. - Pin a command to an agent or model.\n- Set\n`subtask: true`\n\nto force the command into a separate subagent context. - A custom command can override a built-in command with the same name.\n\nExample:\n\n```\n---\ndescription: Review one component\nagent: plan\nsubtask: true\n---\n\nReview @$1 for correctness, accessibility, and performance.\nRun the smallest relevant tests and summarize the risks.\n```\n\n### Agents\n\n- Primary agents:\n**Build** and**Plan**. - Switch primary agents with\n`Tab`\n\nor the configured agent-switch key. - Built-in subagents:\n**General**,** Explore**, and** Scout**. - Invoke a subagent directly with\n`@general`\n\n,`@explore`\n\n, or`@scout`\n\n. - Primary agents may also delegate automatically based on subagent descriptions.\n\n### Skills\n\nOpenCode supports `SKILL.md`\n\nskills and searches `.opencode/skills`\n\n, `.claude/skills`\n\n, and `.agents/skills`\n\nat project and user scope. Skills are discovered by metadata and loaded on demand through the native `skill`\n\ntool.\n\n### Context and shell syntax\n\n`@path`\n\nattaches a file or configured reference.`@agent`\n\ncan invoke a subagent.`!command`\n\nruns a shell command and adds the result to context.`$...`\n\nis primarily template syntax inside custom commands, not a first-class end-user skill selector.\n\n### Distinctive strengths\n\n- Provider-agnostic and open source.\n- Very clean Build/Plan split.\n- Custom commands can select model, agent, and subtask behavior.\n- Compatibility with Claude-style and Agent Skills directories makes it easy to share reusable behavior.\n\n## 6.Cursor\n\nCursor is best understood as a family of connected harnesses rather than only an editor. The editor, terminal CLI, Agents Window, Cloud Agents, and Bugbot share rules, skills, MCP configuration, custom subagents, and session context to different degrees.\n\n### Harnesses\n\n| Harness | Best use | Command-language notes |\n|---|---|---|\nCursor editor Agent |\nRepository work with visual diffs, checkpoints, browser context, and IDE tools | Rich `@` context menu; skills and subagents use `/name` ; Plan and Ask modes are built in |\nCursor CLI |\nTerminal-native agent work and automation | Large slash catalog, `@file` /`@folder` , `/shell` , headless mode, sessions, worktrees, and cloud handoff |\nAgents Window |\nRunning several local agents independently | Native worktree isolation, parallel sessions, review, and applying completed work |\nCloud Agents |\nRemote background work from web, desktop, mobile, integrations, or CLI | Runs in isolated remote environments and branches; CLI handoff uses `& prompt` |\nBugbot |\nPull-request review and autofix | Reviews PRs in GitHub or Bitbucket and can launch a Cloud Agent to fix findings |\n\n### High-value Cursor CLI commands\n\n| Category | Commands |\n|---|---|\n| Modes | `/plan` , `/ask` , `/debug` , `/run-everything` , `/max-mode` |\n| Session/context | `/clear` , `/resume` , `/fork` , `/summarize` , `/context` , `/rewind` , `/rename` |\n| Display/control | `/vim` , `/line-numbers` , `/show-thinking` , `/status-indicators` , `/logs` |\n| Shell/editor | `/shell` , `/sh` , `/run` , `/open` , `/cursor` , `/setup-terminal` |\n| Models/extensions | `/model` , `/mcp` , `/plugin` , `/config` , `/sandbox` , `/bedrock` |\n| Support/account | `/about` , `/help` , `/feedback` , `/copy-request-id` , `/copy-conversation-id` , `/logout` , `/quit` |\n\nCursor CLI also supports:\n\n`agent -p`\n\nor`agent --print`\n\nfor non-interactive output.`--force`\n\nor`--yolo`\n\nwhen a headless run is allowed to modify files.`--worktree [name]`\n\nor`-w`\n\nfor an isolated worktree.`agent ls`\n\n,`agent resume`\n\n,`--continue`\n\n, and`--resume=<id>`\n\nfor session recovery.`agent acp`\n\nwhen Cursor is used through an ACP-compatible host.\n\n### Context, plans, and worktrees\n\n- Use\n`@file`\n\nor`@folder`\n\nto attach repository context. - The editor also exposes\n`@Docs`\n\n,`@Terminals`\n\n,`@Past Chats`\n\n, Git working-state or branch diffs, and`@Browser`\n\n. - Plan mode is available from\n`Shift+Tab`\n\n,`/plan`\n\n,`--plan`\n\n, or`--mode=plan`\n\n. - Ask mode is available as\n`/ask`\n\nor`--mode=ask`\n\n. - The editor supports\n`/worktree`\n\n,`/best-of-n`\n\n,`/apply-worktree`\n\n, and`/delete-worktree`\n\n; the CLI uses`--worktree`\n\n. - In the CLI,\n`Ctrl+R`\n\nopens a review of the agent's changes.\n\n### Skills and rules\n\n- Skills use\n`SKILL.md`\n\nand can be selected automatically or explicitly as`/skill-name`\n\n. `disable-model-invocation: true`\n\nmakes a skill explicit-only.- Cursor discovers skills from\n`.agents/skills`\n\n,`.cursor/skills`\n\n,`.claude/skills`\n\n, and`.codex/skills`\n\n, including supported user-level locations. `/migrate-to-skills`\n\nconverts older command-style customizations.- Project rules live in\n`.cursor/rules/*.mdc`\n\n; Cursor also supports`AGENTS.md`\n\n, user rules, and team rules. - A manually applied rule can be attached with\n`@rule-name`\n\n;`/create-rule`\n\nhelps create a rule.\n\n### Subagents\n\n- Cursor can delegate automatically based on a subagent's description.\n- Explicit invocation uses\n`/subagent-name`\n\nor a natural-language request. - Multiple independent subagent tasks can run in parallel.\n- Built-in subagents include\n**Explore**,** Bash**, and** Browser**. - Custom agents live in\n`.cursor/agents`\n\n; Cursor also reads compatible`.claude/agents`\n\nand`.codex/agents`\n\nlocations. - Agent definitions can specify a model, read-only behavior, and background execution.\n- Background subagent output is stored under\n`~/.cursor/subagents/`\n\n.\n\n### Cloud handoff and review\n\n- Prefix a Cursor CLI prompt with\n`&`\n\nto send it to a Cloud Agent. `/in-cloud`\n\nsends the next delegated task to a cloud subagent.`/babysit`\n\nmonitors and helps shepherd a pull request.`/review`\n\nand`/review-bugbot`\n\nare documented for Cursor Agent and cursor.com/agents in Cursor 3.7 and later. Cursor's documentation currently says CLI support is coming later.- Bugbot Autofix can create a Cloud Agent task from a review finding.\n\n### Distinctive strengths\n\n- One customization model spans editor, CLI, local agents, and cloud agents.\n- The richest\n`@`\n\ncontext vocabulary in this comparison. - Explicit\n`/skill-name`\n\nand`/subagent-name`\n\nsyntax keeps reusable behavior easy to discover. - Native local worktrees, best-of-N execution, cloud handoff, PR review, and browser-aware work.\n\n## 7.GitHub Copilot: compare the harnesses separately\n\n### Copilot's surfaces are not one command language\n\n### 7.1GitHub Copilot CLI\n\nCopilot CLI now has a large terminal-agent command surface.\n\n| Category | Commands and shortcuts |\n|---|---|\n| Planning/autonomy | `/plan` , `Shift+Tab` for standard/plan/autopilot, `--autopilot` |\n| Parallelism | `/fleet` , `/tasks` , `/agents` , `/subagents` , custom agents |\n| Insights/history | `/chronicle` , `/usage` , `/context` , `/session` , `/resume` |\n| Research/review | `/research` , `/review` , `/security-review` , `/rubber-duck` , `/diff` |\n| GitHub workflow | `/pr` , `/delegate` , `#123` issue/PR context |\n| Skills/plugins | `/skills` , `/plugin` , invoke `/skill-name` |\n| Runtime/config | `/model` , `/mcp` , `/lsp` , `/permissions` , `/sandbox` , `/settings` , `/env` |\n| Scheduling | `/every` , `/after` in experimental mode |\n| Session control | `/compact` , `/clear` , `/ask` , `/worktree` , `/remote` , `/app` , `/ide` |\n| Direct context/shell | `@file` , `#issue-or-PR` , `!command` |\n\n`/fleet`\n\n`/fleet`\n\ndecomposes work and runs independent subtasks in parallel. Copilot can choose custom agents for subtasks, and you can name one in the prompt with `@custom-agent-name`\n\n.\n\n```\n/fleet Implement the API validation, add integration tests, and update the docs.\nUse @test-writer for the test work.\n```\n\n`/chronicle`\n\n`/chronicle`\n\nis Copilot CLI's closest equivalent to Claude Code's `/insights`\n\n.\n\nDocumented subcommands include:\n\n`standup`\n\n- summarize recent work.`tips`\n\n- personalized workflow suggestions.`cost tips`\n\n- token-spend feedback.`search`\n\n- search session history.`improve`\n\n- suggest improvements to custom instructions.`reindex`\n\n- rebuild and synchronize the session store.\n\nThe exact short form shown by `/help`\n\nmay vary as the feature evolves.\n\n#### Skills\n\n- Copilot may automatically select skills.\n- Explicitly request one with\n`/skill-name`\n\n. - Manage them with\n`/skills list`\n\n,`/skills info`\n\n,`/skills add`\n\n,`/skills remove`\n\n, and`/skills reload`\n\n.\n\n### 7.2GitHub Copilot app\n\nThe desktop app is built on the Copilot CLI runtime and is designed for managing multiple local and cloud agent sessions. It shares session-history data with the CLI and can contribute to `/chronicle`\n\nreports.\n\nImportant distinction: runtime sharing does not guarantee that every CLI slash command is exposed identically in every app composer. Use the app's live `/`\n\nmenu as the authoritative list.\n\nBest fit:\n\n- Multiple concurrent agent sessions.\n- Visual session management.\n- GitHub-native repository, branch, CI, issue, and PR workflows.\n- MCP servers configured from repositories, CLI configuration, or app settings.\n\n### 7.3Copilot Chat on GitHub.com\n\nGitHub.com Chat uses a lighter command language:\n\n`/clear`\n\n,`/delete`\n\n,`/new`\n\n,`/rename`\n\n.`@`\n\nattaches discussions, extensions, files, issues, pull requests, and repositories.- GitHub MCP skills can perform repository operations from natural language.\n\nThis is a GitHub-object chat surface, not a local terminal harness.\n\n### 7.4Copilot Chat in VS Code\n\nCommon slash commands:\n\n`/clear`\n\n, `/explain`\n\n, `/fix`\n\n, `/fixTestFailure`\n\n, `/help`\n\n, `/new`\n\n, `/tests`\n\nCommon chat variables:\n\n`#block`\n\n, `#class`\n\n, `#comment`\n\n, `#file`\n\n, `#function`\n\n, `#line`\n\n, `#path`\n\n, `#project`\n\n, `#selection`\n\n, `#sym`\n\nCommon chat participants:\n\n`@azure`\n\n, `@github`\n\n, `@terminal`\n\n, `@vscode`\n\n, `@workspace`\n\nUseful query patterns:\n\n```\n@workspace Explain the authentication flow and identify the trust boundaries.\n#file:src/auth.ts /tests Generate tests for the token refresh edge cases.\n@github #web What is the current supported Node.js LTS, and what changes would\nthis repository need to upgrade?\n```\n\nPrompt files under `.github/prompts/*.prompt.md`\n\ncreate reusable `/prompt-name`\n\ncommands in supported IDEs.\n\n### 7.5Copilot cloud agent\n\nCopilot cloud agent runs independently in a GitHub-hosted environment, works on a branch, and can create a pull request. It can be started from GitHub, mobile, IDEs, API, GitHub CLI, the GitHub MCP Server, and external work-management integrations.\n\nDo not confuse:\n\n**IDE agent mode:** acts in the local workspace.**Copilot cloud agent:** acts in a managed GitHub environment.**Copilot CLI:** acts locally unless explicitly delegated or started with a cloud mode.\n\n## 8. Query translation guide\n\n| What you want | Codex | Claude Code | OpenCode | Cursor | Copilot CLI | Copilot VS Code |\n|---|---|---|---|---|---|---|\n| Explain one file | `Explain @src/auth.ts` |\n`Explain @src/auth.ts` |\n`Explain @src/auth.ts` |\n`Explain @src/auth.ts` |\n`Explain @src/auth.ts` |\n`#file:src/auth.ts /explain` |\n| Plan first | `/plan Add passkeys` |\n`/plan Add passkeys` |\nSwitch to Plan, then ask | `/plan Add passkeys` or select Plan mode |\n`/plan Add passkeys` |\nSelect Plan mode |\n| Run three reviews in parallel | Ask to spawn three subagents | Ask for parallel subagents or use a workflow | Mention subagents or create `subtask: true` commands |\nAsk for parallel subagents or use `/best-of-n` |\n`/fleet ...` |\nAsk Agent mode to delegate/subagent if available |\n| Invoke a design skill | `$frontend-design ...` |\n`/frontend-design ...` |\nAsk the agent to load the skill, or wrap it in a custom command | `/frontend-design ...` |\n`/frontend-design ...` |\nUse the Agent Skill or a `/prompt-file` |\n| Review current changes | `/review` , then `/diff` |\n`/code-review` or `/review` |\nUse Plan/review agent or custom `/review` |\nCLI `Ctrl+R` ; Agent `/review` or `/review-bugbot` |\n`/review` , `/diff` |\nUse code review/agent actions |\n| Get usage coaching | `/usage` |\n`/insights` and `/usage` |\nNo built-in equivalent | `/context` ; team analytics are separate |\n`/chronicle tips` , `/chronicle cost tips` |\nQuery synced sessions or use usage views |\n| Run a shell command | `!npm test` |\n`!npm test` |\n`!npm test` |\n`/shell npm test` |\n`!npm test` |\nUse terminal participant/tooling |\n| Start fresh | `/new` |\n`/clear` |\n`/new` |\n`/clear` or `/new` |\n`/new` |\n`/clear` |\n\n## 9. Recommendations by working style\n\n| Working style | Best fit from this comparison | Why |\n|---|---|---|\n| Terminal power user who wants the broadest packaged workflow catalog | Claude Code | Deep command catalog, `/insights` , `/batch` , review and workflow commands |\n| User who wants a persistent objective and a strong desktop multi-thread workspace | Codex | `/goal` , parallel app threads, worktrees, automations, artifacts |\n| Open-source, provider-agnostic user who likes building custom commands | OpenCode | Flexible command templates, agent selection, subtask routing, provider choice |\n| IDE-first user who wants rich context, local parallelism, and easy cloud handoff | Cursor | Deep `@` context, skills and subagents, worktrees, Cloud Agents, Bugbot |\n| GitHub-centered user who wants issues, PRs, session history, parallel workers, and IDE integration | GitHub Copilot | `/fleet` , `/chronicle` , `/pr` , cloud agent, GitHub object context |\n| User who wants a consistent Agent Skills investment across tools | All five, with caveats | All support `SKILL.md` ; invocation syntax and supported extensions differ |\n\n## 10. Other coding agents and harnesses to add later\n\nThis is a candidate backlog, not an exhaustive market inventory.\n\n### Terminal-first or terminal-capable\n\n**Google Antigravity CLI**- Google's consumer successor to Gemini CLI, sharing an agent harness with Antigravity 2.0, IDE, and SDK surfaces** Aider****Amp****Cline CLI****Continue CLI****Crush****Plandex****Pi Coding Agent****Goose****Junie CLI****Kiro CLI****Factory Droid**\n\n### IDE/editor-first\n\n**Devin Desktop / Cascade**(the product formerly known as Windsurf)** Cline****Roo Code****Continue****JetBrains Junie****Kiro IDE****Zed Agent Panel****Augment Code****Replit Agent**\n\n### Cloud/issue-to-PR and engineering-agent platforms\n\n**Devin****OpenHands****Factory****GitLab Duo Agent Platform****Replit Agent****Augment Remote Agents****Warp cloud agents / agent control plane**\n\n### Useful adjacent harness hosts\n\n**Warp**- can run its own agent and host external CLI agents.** Zed with ACP**- editor host for external agents.** VS Code Agent mode**- a harness layer that can host Copilot and third-party agents.\n\n### Recommended next comparison rows\n\nFor the next revision, the most useful additions would be:\n\n**Google Antigravity**- compare its CLI, IDE, Antigravity 2.0 desktop command center, SDK, shared multi-agent harness, skills, hooks, plugins, and asynchronous subagents.**Cline and Roo Code**- model-agnostic VS Code agents with modes, rules, MCP, and skills.** Aider**- mature terminal command language and Git-centered workflow.** Kiro and Junie**- spec-driven and JetBrains-native agent approaches.** Devin, Factory, and OpenHands**- cloud/autonomous engineering-agent category.\n\n## 11. Maintenance protocol\n\nWhen updating this report:\n\n- Record the verification date at the top.\n- Re-check each product's official command reference and live\n`/`\n\nmenu. - Separate stable, preview, and experimental commands.\n- Keep GitHub Copilot surfaces in separate rows.\n- Keep Cursor editor, CLI, Agents Window, Cloud Agents, and Bugbot distinctions explicit.\n- Verify whether a command is a built-in command, a bundled skill, a custom skill, an MCP prompt, or a UI action.\n- Re-check the meanings of\n`/`\n\n,`@`\n\n,`$`\n\n,`#`\n\n, and`!`\n\n. - Update the “other agents” backlog for renames, retirements, and replacements.\n- Add a short change-log entry below.\n\n### Change log\n\n| Date | Change |\n|---|---|\n| 2026-06-20 | Replaced the contents hamburger/close pair with restrained sidebar-state icons; moved the divider beneath the Contents heading; and excluded the document title from scroll-spy highlighting. |\n| 2026-06-20 | Added a reusable content-aware SVG file icon that synchronizes its paper, ink, and accent colors with the active document theme and palette. |\n| 2026-06-20 | Added local variable-font delivery, responsive WebP hero images with intrinsic dimensions, and a Brotli/Gzip static audit server following a Lighthouse performance pass. |\n| 2026-06-20 | Moved the contents, light/dark, and color-palette controls into circular document-corner controls; added a desktop contents hide/show state with a menu-to-close icon transition. |\n| 2026-06-19 | Added a generated `llm.txt` Markdown representation, a canonical `llms.txt` discovery index, HTML alternate-resource metadata, and discoverable footer links. |\n| 2026-06-19 | Restored the original compact command-grammar hero, removed its chroma-key background and outer canvas padding, simplified heading typography, and rebuilt the Open Graph card around three text levels and a fading right-side illustration. |\n| 2026-06-19 | Refined and moved the hero illustration above the title, added mobile-specific diagram compositions, and added Open Graph and Twitter sharing metadata with a dedicated 1200 × 630 social card. |\n| 2026-06-19 | Added an opening line illustration and SVG diagrams for the four-layer request model, sigil portability, and GitHub Copilot harness boundaries. |\n| 2026-06-19 | Replaced the consumer-facing Gemini CLI backlog entry with Google Antigravity. Consumer Gemini CLI service ended June 18, 2026; enterprise and paid API-key access remain supported. |\n| 2026-06-19 | Added Cursor editor, CLI, Agents Window, Cloud Agents, Bugbot, skills, subagents, rules, worktrees, command translations, and official sources. |\n| 2026-06-19 | Initial report. Added Codex, Claude Code, OpenCode, Copilot CLI, Copilot app, GitHub.com Chat, VS Code Chat, cloud-agent distinctions, sigil comparison, and future-agent backlog. |\n\n## Sources\n\nAll core comparisons use official documentation current on June 19, 2026.\n\n### OpenAI Codex\n\n[Codex CLI slash commands](https://developers.openai.com/codex/cli/slash-commands)[Codex app commands](https://developers.openai.com/codex/app/commands)[Codex IDE slash commands](https://developers.openai.com/codex/ide/slash-commands)[Codex Agent Skills](https://developers.openai.com/codex/skills)[Codex subagents](https://developers.openai.com/codex/subagents)[Codex CLI features](https://developers.openai.com/codex/cli/features)[Codex app features](https://developers.openai.com/codex/app/features)\n\n### Claude Code\n\n[Claude Code commands](https://code.claude.com/docs/en/commands)[Claude Code skills](https://code.claude.com/docs/en/skills)[Claude Code subagents](https://code.claude.com/docs/en/sub-agents)[Claude Code parallel agents](https://code.claude.com/docs/en/agents)[Claude Code interactive mode](https://code.claude.com/docs/en/interactive-mode)[Claude Code platforms](https://code.claude.com/docs/en/platforms)\n\n### OpenCode\n\n### Cursor\n\n[Cursor Agent overview](https://cursor.com/docs/agent/overview)[Cursor prompting and](https://cursor.com/docs/agent/prompting)`@`\n\ncontext[Cursor Plan mode](https://cursor.com/docs/agent/plan-mode)[Cursor CLI overview](https://cursor.com/docs/cli/overview)[Using Cursor CLI](https://cursor.com/docs/cli/using)[Cursor CLI slash-command reference](https://cursor.com/docs/cli/reference/slash-commands)[Cursor Agent Skills](https://cursor.com/docs/skills)[Cursor subagents](https://cursor.com/docs/subagents)[Cursor Cloud Agents](https://cursor.com/docs/cloud-agent)[Cursor rules](https://cursor.com/docs/rules)[Cursor Bugbot](https://cursor.com/docs/bugbot)[Cursor worktrees](https://cursor.com/docs/configuration/worktrees)\n\n### GitHub Copilot\n\n[Copilot CLI command reference](https://docs.github.com/en/copilot/reference/copilot-cli-reference/cli-command-reference)[Copilot Chat cheat sheet](https://docs.github.com/en/copilot/reference/chat-cheat-sheet)[Copilot CLI fleet](https://docs.github.com/en/copilot/concepts/agents/copilot-cli/fleet)[Copilot CLI session data and Chronicle](https://docs.github.com/en/copilot/how-tos/copilot-cli/use-copilot-cli/chronicle)[Copilot CLI Agent Skills](https://docs.github.com/en/copilot/how-tos/copilot-cli/customize-copilot/add-skills)[GitHub Copilot app](https://docs.github.com/en/copilot/concepts/agents/github-copilot-app)[Copilot cloud agent](https://docs.github.com/en/copilot/concepts/agents/cloud-agent/about-cloud-agent)[Copilot customization cheat sheet](https://docs.github.com/en/copilot/reference/customization-cheat-sheet)", "url": "https://wpnews.pro/news/the-grammar-of-coding-agents", "canonical_source": "https://grammar-of-coding-agents.pages.dev", "published_at": "2026-06-21 10:16:43+00:00", "updated_at": "2026-06-21 10:37:08.232350+00:00", "lang": "en", "topics": ["ai-tools", "developer-tools", "large-language-models", "generative-ai", "ai-agents"], "entities": ["OpenAI Codex", "Claude Code", "OpenCode", "Cursor", "GitHub Copilot", "OpenAI", "Anthropic", "GitHub"], "alternates": {"html": "https://wpnews.pro/news/the-grammar-of-coding-agents", "markdown": "https://wpnews.pro/news/the-grammar-of-coding-agents.md", "text": "https://wpnews.pro/news/the-grammar-of-coding-agents.txt", "jsonld": "https://wpnews.pro/news/the-grammar-of-coding-agents.jsonld"}}