Supported harnesses

HarnessTap registers 43 agent harnesses today. Each harness declares which resource types it can scan, compose in plugins, and materialize on disk, plus...

HarnessTap registers 43 agent harnesses today. Each harness declares which resource types it can scan, compose in plugins, and materialize on disk, plus default project and global paths. The registry in src/platforms/registry.ts is the source of truth; ht harness list prints the same set at runtime.

For portability caveats (hooks with ${*_PLUGIN_ROOT}, OpenCode server plugins, instruction-only skill emission, and mirror warnings), see Portability limits.

Resource types

HarnessTap separates context-side resources (what the model sees) from environment-side resources (how it runs).

Context-side material resources

These are the atomic resources you scan, curate, attach to plugins, and re-emit during apply:

TypeDescription
instructionsAlways-on context files (AGENTS.md, CLAUDE.md, .windsurfrules, .github/copilot-instructions.md, …)
skillsSKILL.md trees under harness-native skill directories
rulesPath-scoped or always-on rules (.cursor/rules/*.mdc, .claude/rules/, .kiro/steering/, …)
mcpMCP server definitions from harness MCP config files
permissionsAllow/deny patterns (Claude Code and Codex settings)
hooksHook event definitions from settings or hooks.json
agentsAgent manifest files under harness agents/ directories
commandsStatic command definitions (markdown, TOML, or manifest pointers)

Plugin composition adds two attachment kinds on top of material resources:

AttachmentRole
plugin_pinLazy link to a host marketplace or local plugin (plugin_pin:name@marketplace); materializes after resource sync or apply --sync-plugins
pluginNested plugin reference expanded depth-first at apply time

Environment resources

Environments carry how values that override matching plugin resources during apply (home → plugin default → deck active):

TypeDescription
env_varPlain environment variable key/value pairs
model_configDefault model and provider selection
permissionPermission allow/deny overrides
secret_refIndirection to secrets (keychain, env, file) — never embedded in exports

Only harnesses whose registry entry includes env_vars, model_config, and/or permissions and whose native serializer implements those surfaces receive environment values on disk. Today that means Claude Code and Codex for the full environment trio.

All harnesses whose serializer emits MCP config still receive MCP ${VAR} substitution from the environment cascade at apply time (tokens in MCP env / args). OAuth sessions managed by the host IDE are outside this path — see Environments — MCP authentication limitations.

HarnessMCP scan/emitFull env emissionMCP ${VAR} substitution
claude-codeYes (.mcp.json)YesYes
codexYes (config.toml)YesYes
cursorYes (.cursor/mcp.json)NoYes
copilot-cli, github-copilotYesNoYes
opencode, goose, grok-build, antigravity, amazon-q, genericYesNoYes
deepseek-harnessYes (~/.dsh/cordis.patch.yml home patch)NoYes
muse-codeYes (~/.config/muse/settings.json mcp_servers, user-settings only)NoYes

Plugin manifest layouts

scan and plugin-source import recognize these on-disk plugin trees (in addition to per-harness project files):

Manifest pathHarness familyImported as
.claude-plugin/plugin.jsonClaude Codeclaude-plugin
.claude-plugin/marketplace.jsonClaude Code (marketplace root)marketplace
.cursor-plugin/plugin.jsonCursorcursor-plugin
.cursor-plugin/marketplace.jsonCursor (marketplace root)marketplace
.codex-plugin/plugin.jsonCodexcodex-plugin
.github/plugin/plugin.jsonGitHub Copilot / Copilot CLIcopilot-plugin
plugin.json / .goose-plugin/plugin.json / .plugin/plugin.jsonGoose (Open Plugins)goose-plugin

Scanning a repo with both harness files and a plugin manifest merges both sources automatically. See Portability limits — dual-mode scan.

Plugin install and sync providers

During apply, HarnessTap can install and sync plugins from host install trees for:

HarnessProviderTypical install location
claude-codeClaude Code marketplace / claude plugin~/.claude/plugins/
cursorInventory + git check/update; marketplace ensure via agent plugin marketplace add (install remains Cursor Customize / /plugin — no agent plugin install)~/.cursor/plugins/
goosegoose plugin install (git-backed Open Plugins)~/.agents/plugins/
copilot-cliCopilot CLI marketplace / copilot plugin~/.copilot/installed-plugins/
deepseek-harnessdsh plugin --profile web add$DSH_HOME/profiles/web/

Claude Code and Copilot CLI home scan also import installed marketplace plugins as plugin_pin library resources (pins only; use resource sync to materialize children). Claude reads ~/.claude/plugins/installed_plugins.json. Copilot CLI reads ~/.copilot/installed-plugins/<marketplace>/<plugin>/ when ~/.copilot/settings.json or that install tree is present.

Other harnesses still benefit from plugin-source scan and resource sync when you point at an install tree or plugin repo, but do not have an automated install provider yet.

Claude plugin pins and marketplace metadata (plugin showclaude block) apply only when materializing to claude-code.

Serializers

TierHarnessesNotes
Nativeclaude-code, codex, cursor, goose, opencode, github-copilot, copilot-cli, gemini-cli, grok-build, deepseek-harness, muse-codeDedicated scan/serialize logic
GenericAll other registered harnessesPath-driven serializer driven by registry projectPaths / globalPaths

Filter native harnesses at the CLI:

ht harness list --supported
ht harness list --format json

Skill emission modes

Most harnesses emit skills to native skill directories. These harnesses declare skillEmission: instruction-only — skills merge into instructions or rules instead of a separate skill tree:

HarnessEmission target
github-copilot.github/copilot-instructions.md
gemini-cliInstruction/rules paths per Gemini serializer
windsurf.windsurf/rules/{name}.md
cline.clinerules/{name}.md (or merged rules file)
kiro.kiro/steering/{name}.md

Cursor additionally honors project cursor_skill_mode (agent-requested, always-on, agents-skills). Inspect with ht harness project status --project . --format json.

Agent / subagent bridging

Harnesses with a delegated subagent model normalize into a shared canonical shape (description, instruction body, model, reasoning_effort, sandbox/readonly flags) via src/services/agent-bridge.ts:

HarnessNative formatApply notes
codex.codex/agents/*.tomldeveloper_instructions, model_reasoning_effort, sandbox_mode
claude-code.claude/agents/*.md + YAMLRichest metadata; extra Claude-only keys preserved in metadata.extra
cursor.cursor/agents/*.md + YAMLreadonly / is_background; sandbox_mode: read-only maps to readonly: true
opencode, github-copilot, copilot-cli, grok-build*/agents/*.mdMarkdown with optional frontmatter
generic harnesses with agents: paths*.mdSame as OpenCode/Copilot emission

Plugin import scans agents/*.md and agents/*.toml (Codex packs). Cross-harness apply re-emits valid native files per target harness.

Harness reference

Legend for the Resources column: instr instructions · skill skills · rule rules · mcp MCP · perm permissions · hook hooks · agent agents · cmd commands · env env vars · model model config.

IDNameSerializerResourcesSkill emissionPlugin install
claude-codeClaude CodeNativeinstr, skill, rule, mcp, perm, hook, agent, cmd, env, modelNative skillsYes
codexCodexNativeinstr, skill, rule, mcp, perm, hook, agent, env, modelNative skills
cursorCursorNativeinstr, skill, rule, mcp, hook, agentNative skillsYes
opencodeOpenCodeNativeinstr, skill, mcp, agent, cmdNative skills
github-copilotGitHub CopilotNativeinstr, skill, mcp, agentInstruction-only
copilot-cliCopilot CLINativeinstr, skill, mcp, agentNative skills
gemini-cliGemini CLINativeinstr, skill, cmdInstruction-only
antigravityAntigravityGenericinstr, skill, rule, mcp, cmdNative skills
amazon-qAmazon Q DeveloperGenericinstr, rule, mcp
windsurfWindsurfGenericinstr, skill, rule, mcpInstruction-only
clineClineGenericinstr, skill, rule, mcpInstruction-only
rooRoo CodeGenericinstr, skill, rule, mcpNative skills
continueContinueGenericinstr, skill, mcpNative skills
gooseGooseNativeinstr, skill, mcp, hook, cmdNative skillsYes
traeTraeGenericinstr, skill, rule, mcpNative skills
openhandsOpenHandsGenericinstr, skill, mcpNative skills
kiroKiroGenericinstr, skill, ruleInstruction-only
warpWarpGenericinstr, skillNative skills
piPiGenericinstr, skillNative skills
aiderAiderGenericinstr
zedZedGenericinstr, skillNative skills
devinDevinGenericinstr, skillNative skills
julesJulesGenericinstr, skillNative skills
codySourcegraph CodyGenericinstr, mcp
grok-buildGrok BuildNativeinstr, skill, mcp, perm, hook, agent, cmd, modelNative skills
deepseek-harnessDeepSeek HarnessNativeinstr, skill, mcp, perm, hook, agent, modelNative skillsYes
muse-codeMuse CodeNativeinstr, skill, mcp, hookNative skills
ampAmpGenericinstr, skillNative skills
kiloKilo CodeGenericinstr, skillNative skills
augmentAugmentGenericinstr, skillNative skills
firebenderFirebenderGenericinstr, skillNative skills
junieJunieGenericinstr, skillNative skills
zencoderZencoderGenericinstr, skillNative skills
deepagentsDeep AgentsGenericinstr, skillNative skills
qwen-codeQwen CodeGenericinstr, skillNative skills
crushCrushGenericinstr, skillNative skills
droidDroidGenericinstr, skillNative skills
codebuddyCodeBuddyGenericinstr, skillNative skills
muxMuxGenericinstr, skillNative skills
kodeKodeGenericinstr, skillNative skills
command-codeCommand CodeGenericinstr, skillNative skills
cortexCortex CodeGenericinstr, skillNative skills
neovateNeovateGenericinstr, skillNative skills

On-disk paths (selected harnesses)

These are the primary project paths HarnessTap scans and writes. Global paths follow the same keys under each harness home directory (see ht harness list --format json).

HarnessInstructionsSkillsRulesMCPAgentsCommandsSettings
claude-codeCLAUDE.md.claude/skills/.claude/rules/.mcp.json.claude/agents/.claude/commands/.claude/settings.json
codexAGENTS.md.agents/skills/.codex/config.toml.codex/agents/.codex/config.toml
cursorAGENTS.md (+ legacy .cursorrules).agents/skills/.cursor/rules/.cursor/mcp.json.cursor/agents/
opencodeAGENTS.md.opencode/skills/opencode.json.opencode/agents/.opencode/commands/
github-copilot.github/copilot-instructions.md.agents/skills/(global ~/.copilot/mcp-config.json).github/agents/
copilot-cliAGENTS.md.agents/skills/.copilot/mcp-config.json.github/agents/
gemini-cliAGENTS.md.agents/skills/commands/gemini-extension.json
antigravityAGENTS.md (+ GEMINI.md).agents/skills/.agents/rules/.agents/mcp_config.json.agents/workflows/
amazon-qAmazonQ.md (+ AGENTS.md).amazonq/rules/.amazonq/mcp.json
windsurf.windsurfrules.agents/skills/.windsurf/rules/(global MCP config)
rooAGENTS.md.roo/skills/.roomodes.roo/mcp.json
gooseAGENTS.md, .goosehints.agents/skills/ (+ .goose/skills/).config/goose/config.yaml (project), ~/.config/goose/config.yaml (global)recipes/*.yaml~/.config/goose/config.yaml
aiderCONVENTIONS.md (+ AGENTS.md).aider.conf.yml
zedAGENTS.md (+ .rules).agents/skills/
devinAGENTS.md (+ AGENTS.local.md).agents/skills/.devin/config.json
grok-buildAGENTS.md (+ AGENT.md).grok/skills/.grok/config.toml.grok/agents/.agents/commands/.grok/config.toml
deepseek-harnessAGENTS.md (+ CLAUDE.md).dsh/skills/~/.dsh/cordis.patch.yml (global)~/.dsh/.agent-presets/~/.dsh/settings.yaml
muse-codeAGENTS.md (also reads CLAUDE.md, .agents/AGENTS.md, .claude/CLAUDE.md).agents/skills/~/.config/muse/settings.json (mcp_servers, global only)~/.config/muse/settings.json
codyAGENTS.md(global ~/.config/sourcegraph/cody.json)cody.json

Cursor global user skills live under ~/.cursor/skills/. Cursor also maintains app-managed built-ins under ~/.cursor/skills-cursor/ — HarnessTap inventories those on profile status / apply-preview (host_managed.cursor) but never imports or applies them.

Copilot CLI home state lives under ~/.copilot/. Detection uses ~/.copilot/settings.json and ~/.copilot/installed-plugins/ in addition to ~/.copilot/skills/ and ~/.copilot/mcp-config.json. Installed plugins are imported as pins from ~/.copilot/installed-plugins/<marketplace>/<plugin>/.

Generic harnesses in the skills-only tier use harness-specific skill roots such as .kilocode/skills/, .crush/skills/, or .factory/skills/ with AGENTS.md instructions — see the registry for the full list.

Goose context engineering

HarnessTap maps Goose context engineering surfaces as follows:

Goose featureHarnessTap support
goosehints (.goosehints, nested, global ~/.config/goose/.goosehints)Scanned and emitted as instruction resources
Agent skills (.agents/skills/, legacy .goose/skills/)Native skill resources
Open Plugins (plugin.json, .goose-plugin/, .agents/plugins/)Plugin-source import + goose plugin install sync
Hooks (plugin hooks/hooks.json)hook resources; apply emits .agents/plugins/harnesstap-plugin/
MCP extensions (config.yaml extensions:)mcp_server resources via native serializer
Recipes (recipes/*.yaml)command resources
Subagents, plan mode, prompt templates, MOIM, memory extensionRuntime-only — not plugin resources (see portability limits)

Antigravity notes

Antigravity IDE, Antigravity CLI (agy), and AGY share the same workspace layout under .agents/. Workflows (.agents/workflows/*.md) map to HarnessTap commands. Global config lives under ~/.gemini/ (GEMINI.md, config/mcp_config.json, skills/, config/global_workflows/). This is distinct from the thinner gemini-cli entry.

Grok Build notes

Grok Build’s native layout under .grok/ maps as follows:

Grok surfaceHarnessTap support
AGENTS.md (AGENT.md, Agents.md)instruction resources
Skills (.grok/skills/, ~/.grok/skills/, also ~/.agents/skills/)Native skill resources
Agents (.grok/agents/*.md)agent resources
Hooks (.grok/hooks/*.json)hook resources; apply emits .grok/hooks/harnesstap.json
MCP ([mcp_servers] in .grok/config.toml)mcp_server resources
Permissions ([permission] allow/deny/ask or rules)permission resources
Commands (.agents/commands/, ~/.agents/commands/)command resources (skills also appear as slash commands at runtime)
Default model ([models] default in ~/.grok/config.toml)model_config on global apply only — project .grok/config.toml cannot carry [models]
Personas / roles / plugins / sandbox.tomlRuntime/config-only — not plugin resources today

DeepSeek Harness notes

DeepSeek Harness is a developer preview; the Cordis patch schema can change. Home files live under $DSH_HOME if set, otherwise ~/.dsh. Detection uses distinctive .dsh/skills/ or .dsh/hooks/ — not AGENTS.md alone.

DeepSeek surfaceHarnessTap support
AGENTS.md (CLAUDE.md)instruction resources
Skills (.dsh/skills/)Native skill resources
MCP ($DSH_HOME/cordis.patch.yml home patch)Live MCP is global only — not project
Hooks ($DSH_HOME/cordis.patch.yml configPath)Live hooks are $DSH_HOME/hooks/harnesstap.json (Claude command subset via @deepseek-ai/dsh-hooks-claude-code); project .dsh/hooks/ is scanned and written but not wired into Cordis
Agent presets ($DSH_HOME/.agent-presets/)Persona-only user presets — not a copy of shipped standard
PermissionsNamed presets workspace-write and danger-full-access only
Plugin installdsh plugin --profile web add ($DSH_HOME/profiles/web/) — web profile only

Muse Code notes

Muse Code (Meta) user settings live at ~/.config/muse/settings.json. The file must include "schema_version": 1 whenever it exists. HarnessTap merges mcp_servers and user hooks into that file and leaves model defaults, TUI, runtime_capabilities, telemetry, and managed_hooks_path alone. MCP is user-settings only — there is no documented project MCP path.

Detection uses distinctive .muse/hooks.json and/or ~/.config/muse/ (settings.json or the config directory). Shared AGENTS.md / .agents/skills/ alone do not count as Muse. muse init only writes AGENTS.md; a Muse-only machine is still detected from the home config directory.

Muse surfaceHarnessTap support
AGENTS.md (load walk: AGENTS.md, CLAUDE.md, .agents/AGENTS.md, .claude/CLAUDE.md)instruction resources; apply emits AGENTS.md
Skills (<repo>/.agents/skills/, also discovers .claude/skills and .codex/skills)Native skill resources; apply writes .agents/skills/ only (no double-emit to Claude/Codex trees)
Project hooks (<project>/.muse/hooks.json)hook resources
User hooks (hooks in ~/.config/muse/settings.json)Merged into user settings
MCP (mcp_servers in user settings; stdio / streamable_http)mcp_server resources on global apply only
Memory (.agents/memory/)Skipped — not emitted day one
Workflows / observer agentsRuntime-only — out of scope
ht harness list                          # all registered harnesses
ht harness list --supported              # native serializers only
ht harness status                        # global main + alias selection
ht harness project status --project .    # per-project harness prefs + cursor_skill_mode
ht scan . --dry-run              # detect harnesses and resources in a repo
ht environment list                      # named environments (env vars, models, permissions)

See also