Supported harnesses

ai-rulez has 52 built-in presets, one per AI coding harness. A preset writes the harness’s native layout: its file names, frontmatter, directories and settings documents. List the presets you use in config.toml; for a harness that is not listed, write a provider-backed custom preset.

.ai-rulez/config.toml

presets = ["claude", "cursor", "codex", "gemini"]

ai-rulez generate

With one path-scoped rule, one always-on rule, a skill, an agent, a command and an MCP server, those four presets write:

AGENTS.md                         # every rule for codex and gemini (the path-scoped one inlined); imported by CLAUDE.md
CLAUDE.md                         # @AGENTS.md
.mcp.json                         # MCP servers (Claude Code reads the root file)
.claude/settings.json             # the default [session_start] hook
.claude/rules/api.md              # path-scoped rule
.claude/skills/deploy/SKILL.md
.claude/skills/ship/SKILL.md      # the command, as a user-invocable skill
.claude/agents/reviewer.md
.cursor/rules/api.mdc
.cursor/commands/ship.md
.cursor/agents/reviewer.md
.cursor/mcp.json
.agents/skills/deploy/SKILL.md    # shared by codex, cursor and gemini
.agents/skills/ship/SKILL.md      # the command, as a codex skill
.agents/skills/ship/agents/openai.yaml
.codex/agents/reviewer.toml
.codex/config.toml                # MCP servers
.gemini/agents/reviewer.md
.gemini/settings.json             # MCP servers

Codex and Gemini CLI have no rules folder, so the path-scoped rule is also inlined into AGENTS.md, which Claude Code and Cursor read too; generate warns about it and suggests [rules] scoped_fallback = "skill".

ai-rulez init lists every built-in preset in a comment of the new config.toml. Three more preset names are not harnesses: mcp (only the root .mcp.json), okf and llms-txt.

Feature matrix

yes means the preset writes that output at project level, - that it does not (the reason is under Skipped features).

• Rules: a native rules folder. - means rules are inlined into the root instructions file (letta has none, so it gets no rules).
• Commands: slash commands. Where the harness has no command folder, a command is written as a skill (claude, codex, antigravity, devin, warp) or as a prompt or workflow file (copilot, pi, cline, kiro).
• Hooks and Perms: yes renders into a project file; user renders only with generate --user, because the harness reads them from the user configuration only. See Hooks and settings and Permissions.
• Checks: code-review guidelines.
• copilot-cli permissions cover only WebFetch URL rules, and gitlab-duo hooks only the SessionStart event.
• User: supported by generate --user; see User-level configuration.

Preset	Harness	Rules	Skills	Agents	Commands	MCP	Hooks	Perms	Checks	User
aiassistant	JetBrains AI Assistant	yes	yes	-	-	yes	-	-	-	-
amp	Sourcegraph Amp	-	yes	-	-	yes	yes	-	yes	yes
antigravity	Google Antigravity	yes	yes	yes	yes	yes	yes	-	-	yes
augment	Augment Code	yes	yes	yes	yes	yes	yes	yes	yes	yes
baz	Baz	-	yes	yes	-	-	-	-	-	-
bob	IBM Bob	yes	yes	-	yes	yes	yes	-	-	yes
claude	Claude Code	yes	yes	yes	yes	yes	yes	yes	-	yes
cline	Cline	yes	yes	yes	yes	-	yes	-	-	yes
codebuddy	CodeBuddy Code	yes	yes	yes	yes	yes	yes	yes	-	yes
codebuff	Codebuff	-	yes	-	-	yes	-	-	-	-
codewhale	Codewhale	yes	yes	-	yes	yes	-	-	-	yes
codex	OpenAI Codex	-	yes	yes	yes	yes	yes	yes	-	yes
commandcode	Command Code	-	yes	yes	yes	yes	yes	yes	-	yes
copilot	GitHub Copilot	yes	yes	yes	yes	yes	yes	yes	-	yes
copilot-cli	GitHub Copilot CLI	yes	yes	yes	-	yes	yes	yes	-	yes
cortex	Snowflake Cortex Code	-	yes	yes	-	-	yes	-	-	yes
crush	Crush	-	yes	-	-	yes	yes	-	-	yes
cursor	Cursor	yes	yes	yes	yes	yes	yes	yes	yes	yes
deepagents	Deep Agents Code	-	yes	yes	-	yes	yes	-	-	yes
devin	Devin CLI	yes	yes	yes	yes	yes	yes	yes	-	yes
dsh	DeepSeek Harness	-	yes	-	-	-	-	-	-	yes
factory	Factory Droid	-	yes	yes	yes	yes	yes	-	yes	yes
gemini	Gemini CLI	-	yes	yes	-	yes	yes	yes	-	yes
gitlab-duo	GitLab Duo CLI	-	-	-	yes	yes	yes	-	yes	yes
goose	goose	-	yes	yes	-	yes	yes	-	-	yes
grok	Grok Build CLI	yes	yes	yes	yes	yes	yes	yes	-	yes
hermes	Hermes Agent	-	agents_md	-	-	-	user	user	-	yes
junie	JetBrains Junie	yes	yes	yes	yes	yes	user	-	-	yes
kilo	Kilo Code	yes	yes	yes	yes	yes	yes	yes	yes	yes
kimi	Kimi Code	-	yes	yes	-	yes	user	user	-	yes
kiro	Kiro	yes	yes	yes	yes	yes	yes	-	-	yes
letta	Letta Code	-	yes	yes	yes	-	yes	yes	-	yes
mimocode	MiMo Code	-	yes	yes	yes	yes	yes	yes	-	yes
muse	Meta Muse Code	-	yes	-	-	-	-	-	-	yes
omp	oh-my-pi	yes	yes	yes	yes	yes	-	yes	-	yes
openclaw	OpenClaw	-	-	-	-	-	-	-	-	yes
opencode	OpenCode	-	yes	yes	yes	yes	yes	yes	-	yes
pi	pi	-	yes	yes	yes	yes	yes	-	-	yes
poolside	Poolside Pool	-	yes	-	-	yes	yes	yes	-	yes
qoder	Qoder	yes	yes	yes	yes	yes	yes	yes	-	yes
qwen	Qwen Code	yes	yes	yes	yes	yes	yes	yes	yes	yes
reasonix	Reasonix	-	yes	yes	yes	yes	yes	-	-	yes
replit	Replit Agent	-	yes	-	-	-	-	-	-	-
rovodev	Atlassian Rovo Dev	-	yes	yes	-	-	-	-	yes	yes
takt	Takt	yes	yes	yes	yes	-	-	-	-	yes
trae	Trae	yes	yes	-	-	yes	-	-	-	yes
vibe	Mistral Vibe	-	yes	-	-	yes	yes	yes	-	yes
warp	Warp	-	yes	-	yes	yes	-	-	-	yes
xum	Xum	-	yes	yes	-	yes	-	-	-	-
zcode	ZCode	-	yes	yes	yes	yes	user	-	-	yes
zed	Zed	-	yes	-	-	yes	-	user	-	yes
zoocode	Zoo Code	-	yes	-	yes	yes	-	yes	-	yes

What each preset writes

The paths below are the project-level outputs with the default agents_md = true: always-on rules and context go into one shared AGENTS.md and skills into one shared .agents/skills/, which every preset that reads them uses instead of its own copy (see AGENTS.md and .agents/skills). <id> is the item’s name. Every project also gets the root .mcp.json, whatever its presets: it carries the ai-rulez MCP server ([mcp] self_server, on by default) and any servers you declare.

Instructions, rules and skills

A rules folder holds the path-scoped rules; with agents_md = false it also holds the always-on rules, and the presets marked * write their own root file and skills folder instead (next table).

Preset	Instructions	Rules	Skills
aiassistant	none	.aiassistant/rules/<id>.md (every rule)	.agents/skills/
amp	AGENTS.md	inlined	.agents/skills/
antigravity *	AGENTS.md	.agents/rules/<id>.md	.agents/skills/
augment *	AGENTS.md	.augment/rules/<id>.md	.agents/skills/
baz	AGENTS.md, <dir>/AGENTS.md (Baz)	nested AGENTS.md	.agents/skills/
bob	AGENTS.md	.bob/rules/<id>.md	.bob/skills/
claude	CLAUDE.md (imports @AGENTS.md)	.claude/rules/<id>.md	.claude/skills/
cline *	AGENTS.md	.clinerules/<id>.md	.agents/skills/
codebuddy	CODEBUDDY.md	.codebuddy/rules/<id>.md (every rule)	.codebuddy/skills/
codebuff	AGENTS.md	inlined	.agents/skills/
codewhale	AGENTS.md	.codewhale/rules/<id>.md	.codewhale/skills/
codex	AGENTS.md	inlined	.agents/skills/
commandcode	AGENTS.md	inlined	.commandcode/skills/
copilot, copilot-cli *	AGENTS.md	.github/instructions/<id>.instructions.md	.agents/skills/
cortex *	AGENTS.md	inlined	.agents/skills/
crush *	AGENTS.md	inlined	.agents/skills/
cursor	AGENTS.md	.cursor/rules/<id>.mdc	.agents/skills/
deepagents *	AGENTS.md	inlined	.agents/skills/
devin *	AGENTS.md	.devin/rules/<id>.md	.agents/skills/
dsh *	AGENTS.md	inlined	.agents/skills/
factory	AGENTS.md	inlined	.factory/skills/
gemini *	AGENTS.md	inlined	.agents/skills/
gitlab-duo	.gitlab/duo/chat-rules.md	inlined	none
goose *	AGENTS.md	inlined	.agents/skills/
grok	AGENTS.md	.grok/rules/<id>.md	.grok/skills/
hermes *	AGENTS.md	inlined	.agents/skills/
junie *	AGENTS.md	.junie/rules/<id>.md	.agents/skills/
kilo	AGENTS.md	.kilo/rules/<id>.md	.kilo/skills/
kimi *	AGENTS.md	inlined	.agents/skills/
kiro *	AGENTS.md	.kiro/steering/<id>.md	.agents/skills/
letta	none	not written	.agents/skills/
mimocode	AGENTS.md	inlined	.mimocode/skills/
muse, pi	AGENTS.md	inlined	.agents/skills/
omp *	AGENTS.md	.omp/rules/<id>.md	.agents/skills/
openclaw	AGENTS.md	inlined	none
opencode *	AGENTS.md	inlined	.agents/skills/
poolside *	AGENTS.md	inlined	.agents/skills/
qoder	AGENTS.md	.qoder/rules/<id>.md	.qoder/skills/
qwen *	AGENTS.md	.qwen/rules/<id>.md	.qwen/skills/
reasonix	REASONIX.md	inlined	.reasonix/skills/
replit	replit.md	inlined	.agents/skills/
rovodev *	AGENTS.md	inlined	.agents/skills/
takt	none	.takt/facets/policies/<id>.md (every rule)	.takt/facets/knowledge/<id>.md
trae	none	.trae/rules/<id>.md (every rule)	.trae/skills/
vibe *	AGENTS.md	inlined	.agents/skills/
warp *	AGENTS.md	inlined	.agents/skills/
xum *	AGENTS.md	inlined	.agents/skills/
zcode	AGENTS.md	inlined	.zcode/skills/
zed *	AGENTS.md	inlined	.agents/skills/
zoocode *	AGENTS.md	inlined	.agents/skills/

With agents_md = false, the presets marked * write:

Preset	Instructions	Skills
antigravity	GEMINI.md	.agents/skills/
augment	AGENTS.md	.augment/skills/
cline	none (every rule in .clinerules/)	.cline/skills/
copilot, copilot-cli	.github/copilot-instructions.md	.github/skills/
cortex	AGENTS.md	.cortex/skills/
crush	CRUSH.md	.crush/skills/
deepagents	AGENTS.md	.deepagents/skills/
devin	none (every rule in .devin/rules/)	.devin/skills/
dsh	AGENTS.md	.dsh/skills/
gemini	GEMINI.md	.agents/skills/
goose	.goosehints	.agents/skills/
hermes	.hermes.md	none
junie	AGENTS.md	.junie/skills/
kimi	AGENTS.md	.kimi-code/skills/
kiro	AGENTS.md	.kiro/skills/
omp	AGENTS.md	.omp/skills/
opencode	AGENTS.md	.opencode/skills/
poolside	AGENTS.md	.poolside/skills/
qwen	QWEN.md	.qwen/skills/
rovodev	AGENTS.md	.rovodev/skills/
vibe	AGENTS.md	.vibe/skills/
warp	AGENTS.md	.warp/skills/
xum	AGENTS.md	.xum/skills/
zed	.rules	.agents/skills/
zoocode	AGENTS.md	.roo/skills/

Agents, commands, MCP and settings

Settings files are written only when the project declares hooks or permissions, except .claude/settings.json, which also holds the [session_start] hook that a claude project gets by default; (user) marks a file that only generate --user writes.

Preset	Agents	Commands	MCP	Hooks and permissions
aiassistant	-	-	.ai/mcp/mcp.json	-
amp	-	-	.amp/settings.json	.amp/plugins/ai-rulez-hooks.ts
antigravity	.agents/agents/<id>.md	.agents/skills/<id>/SKILL.md	.agents/mcp_config.json	.agents/hooks.json
augment	.augment/agents/<id>.md	.augment/commands/<id>.md	.augment/settings.json	.augment/settings.json
baz	.claude/agents/<id>.md	-	-	-
bob	-	.bob/commands/<id>.md	.bob/mcp.json	.bob/settings.json
claude	.claude/agents/<id>.md	.claude/skills/<id>/SKILL.md	.mcp.json	.claude/settings.json
cline	.cline/agents/<id>.yaml	.clinerules/workflows/<id>.md	-	.clinerules/hooks/<Event>
codebuddy	.codebuddy/agents/<id>.md	.codebuddy/commands/<id>.md	.mcp.json	.codebuddy/settings.json
codebuff	-	-	.agents/mcp.json	-
codewhale	-	.codewhale/commands/<id>.md	.codewhale/mcp.json	-
codex	.codex/agents/<id>.toml	.agents/skills/<id>/SKILL.md	.codex/config.toml	.codex/hooks.json, .codex/rules/ai-rulez.rules
commandcode	.commandcode/agents/<id>.md	.commandcode/commands/<id>.md	.mcp.json	.commandcode/settings.json
copilot	.github/agents/<id>.agent.md	.github/prompts/<id>.prompt.md	.vscode/mcp.json	.github/hooks/ai-rulez.json, .vscode/settings.json
copilot-cli	.github/agents/<id>.agent.md	-	.github/mcp.json	.github/hooks/ai-rulez.json, .github/copilot/settings.json
cortex	.cortex/agents/<id>.md	-	-	.cortex/settings.json
crush	-	-	crush.json	crush.json
cursor	.cursor/agents/<id>.md	.cursor/commands/<id>.md	.cursor/mcp.json	.cursor/hooks.json, .cursor/cli.json
deepagents	.deepagents/agents/<id>/AGENTS.md	-	.deepagents/.mcp.json	.deepagents/hooks.json
devin	.devin/agents/<id>.md	.agents/skills/<id>/SKILL.md	.devin/mcp_config.json	.devin/hooks.v1.json, .devin/config.json
dsh	-	-	-	-
factory	.factory/droids/<id>.md	.factory/commands/<id>.md	.factory/mcp.json	.factory/hooks.json
gemini	.gemini/agents/<id>.md	-	.gemini/settings.json	.gemini/settings.json
gitlab-duo	-	.agents/commands/<id>.md	.gitlab/duo/mcp.json	.gitlab/duo/hooks.json
goose	.goose/agents/<id>.md	-	.agents/plugins/ai-rulez/.mcp.json	.agents/plugins/ai-rulez/hooks/hooks.json
grok	.grok/agents/<id>.md	.grok/commands/<id>.md	.grok/config.toml	.grok/hooks/ai-rulez.json, .grok/config.toml
hermes	-	-	-	~/.hermes/config.yaml (user)
junie	.junie/agents/<id>.md	.junie/commands/<id>.md	.junie/mcp/mcp.json	~/.junie/config.json (user)
kilo	.kilo/agents/<id>.md	.kilo/commands/<id>.md	kilo.jsonc	.kilo/plugins/ai-rulez-hooks.js, kilo.jsonc
kimi	.kimi-code/agents/<id>.md	-	.kimi-code/mcp.json	~/.kimi-code/config.toml (user)
kiro	.kiro/agents/<id>.md	.kiro/prompts/<id>.md	.kiro/settings/mcp.json	.kiro/hooks/ai-rulez.json
letta	.letta/agents/<id>.md	.commands/<id>.md	-	.letta/settings.json
mimocode	.mimocode/agents/<id>.md	.mimocode/commands/<id>.md	.mimocode/mimocode.jsonc	.mimocode/plugins/ai-rulez-hooks.js, .mimocode/mimocode.jsonc
muse	-	-	-	-
omp	.omp/agents/<id>.md	.omp/commands/<id>.md	.omp/mcp.json	.omp/config.yml
openclaw	-	-	-	-
opencode	.opencode/agents/<id>.md	.opencode/commands/<id>.md	opencode.json	.opencode/plugins/ai-rulez-hooks.js, opencode.json
pi	.pi/agents/<id>.md	.pi/prompts/<id>.md	.pi/mcp.json	.pi/extensions/ai-rulez-hooks.ts
poolside	-	-	.poolside/settings.yaml	.poolside/settings.yaml
qoder	.qoder/agents/<id>.md	.qoder/commands/<id>.md	.mcp.json	.qoder/settings.json
qwen	.qwen/agents/<id>.md	.qwen/commands/<id>.md	.qwen/settings.json	.qwen/settings.json
reasonix	.reasonix/skills/<id>/SKILL.md	.reasonix/commands/<id>.md	.mcp.json	.reasonix/settings.json
replit	-	-	-	-
rovodev	.rovodev/subagents/<id>.md	-	-	-
takt	.takt/facets/personas/<id>.md	.takt/facets/instructions/<id>.md	-	-
trae	-	-	.trae/mcp.json	-
vibe	-	-	.vibe/config.toml	.vibe/hooks.toml, .vibe/config.toml
warp	-	.agents/skills/<id>/SKILL.md	.warp/.mcp.json	-
xum	.xum/agents/<id>.md	-	.xum/mcp.jsonc	-
zcode	.zcode/agents/<id>.md	.zcode/commands/<id>.md	.zcode/config.json	~/.zcode/cli/config.json (user)
zed	-	-	.zed/settings.json	~/.config/zed/settings.json (user)
zoocode	-	.roo/commands/<id>.md	.roo/mcp.json	.vscode/settings.json

Checks are written by eight presets: amp (.agents/checks/<id>.md), augment (.augment/code_review_guidelines.yaml), cursor (.cursor/BUGBOT.md), factory (.factory/skills/review-guidelines/SKILL.md), gitlab-duo (.gitlab/duo/mr-review-instructions.yaml), kilo (REVIEW.md), qwen (.qwen/review-rules.md) and rovodev (.rovodev/.review-agent.md).

User-level paths are listed in User-level configuration.

Cross-cutting behavior

• Shared outputs. Presets that write the same path (AGENTS.md, .agents/skills/, the root .mcp.json) must render identical bytes. generate fails and names the presets when they diverge. See AGENTS.md and .agents/skills.
• Merged documents. Settings files that also hold your own keys (.claude/settings.json, opencode.json, .codex/config.toml, .qwen/settings.json, …) are merged, not replaced: ai-rulez owns only the keys, array elements and map members it writes, and comments in JSONC, TOML and YAML survive. See Settings document merge behavior.
• External plugins. [[enabled_plugins]] enables a plugin from another marketplace for claude (enabledPlugins, extraKnownMarketplaces), codex ([plugins."<id>"]) and opencode (the plugin array). The other presets have no equivalent and skip it. See Enable a plugin from another marketplace.
• MCP environment references. Where the harness expands ${VAR}-style references itself, the reference is written instead of the secret. See Environment references.
• Built-in names. A skill, command or agent named like a harness built-in replaces it, is renamed, or is unreachable, depending on the harness. validate checks every output against the harnesses that load it, including those that only read a shared directory such as .agents/skills/: AR9U0 for a built-in name, AR9U1 for a name the harness refuses to load. See Built-in names of harnesses.
• Tool and model names. Frontmatter tool names are translated to each harness’s own names, and bare Claude model aliases (sonnet, opus, haiku) are dropped where the harness has no such model.

Skipped features

A - in the matrix is a deliberate skip: ai-rulez does not approximate a format the harness does not document.

Instructions, rules, skills and agents

Preset	Skipped	Reason
aiassistant	root file	AI Assistant has no root instructions file; every rule and context item becomes a .aiassistant/rules/ file
amp	agents	Amp has no agent file of its own and does not read the .agents/agents format
baz	rules folder, commands, MCP	Baz reads only instruction files from the default branch; it ignores .claude/commands and does not read .claude/rules. See Baz
bob	agents	Bob’s modes live in an aggregated custom_modes.yaml, not in one file per agent
codebuff	rules, agents, commands	Codebuff documents only AGENTS.md, skills and MCP
codewhale	agents	Subagents are TOML profiles whose strict schema the preset cannot produce
crush	rules folder, agents, commands	Crush has no rules folder, agent or command format
cline	MCP	Cline has no project MCP file
cortex	rules folder, MCP	Cortex reads only AGENTS.md; MCP servers live in ~/.snowflake/cortex/mcp.json
dsh	MCP	Servers live in the home-level cordis.patch.yml
gemini	rules folder, commands	Rules stay in the root file; project commands are not generated (see the plugin runtime for commands/<name>.toml)
gitlab-duo	skills, rules folder	GitLab documents skills only inside plugins; Duo CLI reads one rules file, .gitlab/duo/chat-rules.md
goose	rules folder, commands	goose has no rules folder, and recipes are not slash commands
hermes	skills	Skills come only from the shared .agents/skills that agents_md = true writes
letta	rules, MCP	Letta documents no instructions, rules or MCP file a project can use
muse	MCP	MCP servers live only in the user settings.json
openclaw	everything but the root file	Only AGENTS.md is written
poolside	agents	Subagents are entries of the settings file, which the preset does not generate
replit	rules, agents, commands, MCP	Replit reads replit.md and .agents/skills only
rovodev	commands, MCP	Saved prompts need a prompts.yml manifest; .rovodev/mcp.json is inert until config.yml points at it
takt	MCP	Takt keeps MCP definitions inside workflow YAML
trae	root file	Trae has no root instructions file
vibe	agents	A Vibe agent is a TOML profile plus a separate prompt file
xum	commands	Xum documents no slash-command file
zoocode	agents	Custom modes are one aggregated .roomodes file
zed	agents, commands, rules folder	Zed has no rules folder, subagent or command format; rules go to the root file
amp, codex, commandcode, deepagents, dsh, factory, hermes, kimi, mimocode, muse, opencode, pi, poolside, reasonix, rovodev, vibe, warp, xum, zcode, zoocode	rules folder	The harness has no rules folder (or one it does not auto-load); rules are inlined into the root file
aiassistant, dsh, gitlab-duo, hermes, muse, trae, warp	agents	The harness documents no agent file
aiassistant, amp, copilot-cli, cortex, deepagents, dsh, hermes, kimi, muse, poolside, trae, vibe	commands	The harness documents no command file
hermes	MCP	The harness documents no project MCP file

Hooks

Sixteen presets generate no [[hooks]]: aiassistant, baz, codebuff, codewhale, dsh, muse, omp, openclaw, replit, rovodev, takt, trae, warp, xum, zed and zoocode. generate names them in one warning. junie, zcode, hermes and kimi read hooks from the user configuration only, so they get hooks from generate --user. Events, matchers and timeouts per harness are in Hooks and settings.

Permissions

The harnesses without a documented, committable permission file that can hold the rules get nothing from [permissions], and generate names them in a warning. The reasons are in Permissions. zed, hermes and kimi read permissions from the user configuration only.

User scope

aiassistant, baz, codebuff, replit and xum have no documented user-level location for anything generate --user writes, so they are reported and skipped. User-level MCP servers are never generated.