Does longhorizon-harness auto-invoke its skills?

Not yet. It is indexed on Flowy as a plain plugin. Request auto-invocation on its page and Flowy will route its skills for you as you prompt.

Is longhorizon-harness free?

Yes. Flowy is free and open source, with nothing gated. You can read every skill in full before you install.

Automation

Plugin

longhorizon-harness

The long-horizon computer-use harness. Run AI agents across desktop apps and the CLI for extended periods while preserving task state and making reliable progress on complex workflows. Features fresh-context execution, durable verified state, independent auditing, recoverable progress, and native Claude Code / Codex / OpenClaw integration.

FLOW.md

amap-ml@amap-ml

GitHub

Plugin Signal74/100Solid

maintained99

capable37

documented100

adopted57

Install

$ npx -y skills add AMAP-ML/LongHorizon-Harness --agent claude-code

Run the curl in your terminal, the rest in Claude Code.

Stats

Python JavaScript Node Bash Docker

+7 tags

AWS Linux GitHub Anthropic OpenAI Hugging Face MCP

515

Stars

Forks

Active

Maintenance

Python

Language

MIT

License

2d ago

Last commit

5d ago

Created

Repo: AMAP-ML/LongHorizon-Harness

In this plugin

5 skills

See everything inside

What's inside

Skills

/analyze-taskCheck OSWorld tasks. Validate the evaluation function, verify that the instruction is feasible given the task setup and agent-visible files, inspect setup artifacts when needed, and produce both markdown and structured JSON reports./analyze-trajAnalyze OSWorld-V2 agent trajectory logs and task results to produce actionable insights. Use this skill whenever the user wants to understand agent performance on OSWorld tasks — including analyzing trajectories, reviewing task results, finding error patterns, comparing code vs /migrate-osworld-agentMigrate an agent from upstream OSWorld into this OSWorld-V2 repository, add matching evaluation entrypoints, and verify the integration./setup-osworldProvision and verify an OSWorld-V2 checkout after clone. Use when the user asks for OSWorld-V2 setup, installation, onboarding, AWS provider setup, Docker provider setup, mocked website server setup, GitLab server setup, gated task download, CUA-Harness hybrid experiment setup,/weavebench-cua-reproduceReproduce CUA-Harness experiments on WeaveBench from a GitHub checkout. Use when the user wants an AI coding agent to set up dependencies, download WeaveBench assets, prepare the 120G VM, configure Qwen/Anthropic-compatible APIs, run smoke tests, launch full or subset

README

Read on GitHub

LongHorizon-Harness

Advancing Long-Horizon Agents for Real-World Tasks

Operate the whole computer like a human. Work across desktop apps and the command line for dozens of hours.

No state drift. Verifiable progress. Complex tasks carried through to completion.

Usage · What You Get · How It Works · Results · Project Website · 简体中文

The model determines what an agent can do in one round. LongHorizon-Harness determines whether that work can be verified, preserved, and continued until the task is actually complete.

Works with Claude Code and Codex. One-command install, ready to run.

LongHorizon-Harness is an execution, state-management, and result-verification system for long-horizon tasks. It does not train a new model or replace an existing agent. It runs on top of systems such as Codex and Claude Code, helping agents operate autonomously in real computer environments for extended periods and continuously move complex tasks forward.

✨ News

[2026-08-07] A new, more user-friendly Dashboard is in the works. Stay tuned.
[v0.1.3 · 2026-08-07] Every run now ends with a plain-language reply that answers your task from the verified state alone. Tasks act on the directory you launched from by default, and the console reports each round as it happens.
[2026-08-06] LongHorizon-Harness reaches #1 on the Hugging Face Daily Papers weekly ranking.
[v0.1.2 · 2026-08-06] Adds unified computer-use plugin management, stronger auditor read-only checks and role isolation, reliable process cleanup, and expanded doctor diagnostics. See Manage computer-use plugins.
[2026-08-06] The WeChat group is open. Scan the QR code below to join.

🚀 We’re iterating rapidly. Stay tuned!

Video Demo

https://github.com/user-attachments/assets/ca8b77ce-9220-4d85-a272-b346009b2454

Three roles. One trusted state.

LongHorizon-Harness separates planning, execution, and verification so that one growing context is not responsible for everything.

	Role	One responsibility
🧭	Manager	Maintains the original goal, verified progress, and next step
⚡	Executor	Starts each round with a fresh context and focuses on one clearly defined task
🔍	Auditor	Independently inspects files, interfaces, logs, and tests in the real environment

Only results that pass independent verification enter persistent task state. Even when the context is refreshed, an action fails, or a deliverable does not pass inspection, the system retains previously verified progress and continues from what remains.

Desktop apps and CLI. One continuous task.

LongHorizon-Harness supports both GUI and CLI workflows.

🖥️ Operate the desktop	⌨️ Work in the terminal
🌐 Click, type, scroll, and browse	💻 Write and modify code
📊 Operate spreadsheets	▶️ Run commands and scripts
📄 Edit documents	📦 Install dependencies and environments
🎨 Use design software	🔧 Configure and debug systems
🧊 Operate 3D tools	📁 Process files and data

One task can begin in a browser, move to the command line for data processing, continue in desktop software to produce an artifact, and return to the terminal for validation or debugging. The goal, progress, and evidence remain under the same state-management system throughout.

Any model. Any agent backend.

LongHorizon-Harness is not tied to a specific model or agent backend. Existing models and agents connect through configuration without changing their original workflows.

	Layer	Supported choices
🧠	Models	Claude, GPT, Qwen, and other models exposed by an agent backend
🤖	Agent backends	Claude Code, Codex CLI, and custom `AgentAdapter` implementations
🎛️	Role assignment	The Manager, Executor, and Auditor can each use a different model or backend
🖥️	Execution environments	Local, with a pluggable `Environment` protocol

A lightweight AgentAdapter preserves each agent's native execution loop while LongHorizon-Harness coordinates role boundaries, verified task state, and cross-round progress around it.

Use one model for all three roles, or combine different models and backends to balance quality, speed, and cost.

Hundreds of real tasks. Measured gains.

LongHorizon-Harness is not demonstrated only on a handful of carefully selected success cases.

We ran it on hundreds of complex tasks across GUI, CLI, and mixed computer environments:

Task domain	What the tasks involve
🌐 Web Frontend	Developing, fixing, and validating websites and web applications through browser interaction, developer tools, and code changes
📊 Data Analysis & Visualization	Processing data, producing charts and dashboards, and checking analytical results and visual deliverables
🛠️ Operations & Debugging	Investigating logs, networks, performance, and service failures; configuring, diagnosing, and repairing systems
🎨 Design & Image Processing	Editing visual assets, matching design references, processing images, and verifying final visual quality
🎮 Games & Interaction	Building, operating, and debugging games or interactive applications; checking interaction logic and runtime behavior
📄 Documents & Presentations	Editing documents and slide decks, including content, formatting, references, layout, and final delivery
🧊 Spatial Reasoning	Completing tasks involving spatial relationships, geometry, precise placement, and 3D operations
🖥️ Desktop & System Settings	Operating desktop applications, files, and system settings across multi-application workflows
🔬 Research & Education	Completing literature research, coursework, teaching materials, forms, and research-support workflows
🎬 Creative Production	Producing presentations, video, audio, and other media while coordinating assets across tools
⚙️ Engineering & Computing	Using CAD, EDA, scientific software, development tools, and cloud or DevOps toolchains
🎫 Personal Services	Handling event ticketing, everyday services, games, and visual-search workflows
🏛️ Administration & Compliance	Completing office, legal, policy-sensitive form, institutional, and safety-aware submission workflows
💼 Business & Finance	Handling market analysis, procurement, loans, sales, reimbursements, and cross-application enterprise workflows
🏥 Healthcare	Completing medical quality-control, insurance, immunization, and structured health-form workflows

Same model. Same execution backend. Only the harness changes.

📊 Full benchmark results and experimental settings

Benchmark	Metric	Claude Code	LongHorizon-Harness	Gain
WeaveBench (114 tasks)	PassRate	51.8	80.7	+28.9
WeaveBench	Overall	0.702	0.835	+0.133
OSWorld 2.0 (108 tasks)	Binary	2.8	8.3	3.0×
OSWorld 2.0	Partial	21.5	35.2	+13.7
Terminal-Bench 2.1	Success rate	69.7	77.2	+7.5

All rows use Qwen 3.7-Plus as the backbone and Claude Code as the execution backend.

Full result tables and case trajectories are available on the LongHorizon-Harness project website.

One command. Full visibility.

Installation

Steps 1–2 are once per machine; steps 3–4 are once per project.

Requirements

	Needed for
uv	The recommended isolated install. Skip it if you prefer pip.
Python 3.10 or later	Running the harness. `uv tool install` brings its own; a pip install uses yours.
One agent runtime on `PATH`: `codex` or `claude`	Actually executing the work. Install both if you want to mix them across roles.
Node.js 20 or later	Only the npm-distributed computer-use plugins. Not needed for `codex-computer-use` or CLI-only tasks.

Platform status: Currently tested on macOS. Windows support is included but has not yet been thoroughly tested.

Run lh-harness doctor at any point to check all of the above; see Verify the environment.

1. Install LongHorizon-Harness

uv tool install lh-harness            # or: pip install lh-harness

Upgrade later with uv tool upgrade lh-harness or pip install --upgrade lh-harness.

2. Install a computer-use plugin

Skip this if your tasks never touch the GUI. Otherwise install the one that matches your agent. No plugin is enabled by default, and one install covers every project on the machine.

Using Codex:

lh-harness plugin install codex-computer-use

Using Claude Code, or both agents:

lh-harness plugin install open-computer-use

codex-computer-use is the official plugin bundled with the Codex CLI and only works with Codex. open-computer-use is distributed on npm, needs Node.js 20+, and drives both agents. Both need OS permissions that must be granted by hand on macOS. See Manage computer-use plugins for that, for clawdcursor as a third option, and for how each one is wired.

3. Generate a project configuration

cd /path/to/your/project
lh-harness init

This creates ./.lh-harness/config.toml without replacing an existing file; use lh-harness init --force to regenerate. Open it and adjust the defaults. Every field is documented in Configuration reference.

4. Run a task

TASK="Inspect the current directory and summarize its files."
lh-harness run --task "${TASK}" --agent codex

Explicit CLI arguments such as --agent override the matching values in ./.lh-harness/config.toml for that run; drop them to use the configured defaults.

The agents work in the directory you launched from, so the task acts on your real project. Set workspace or --workspace to point somewhere else. ./.lh-harness/ itself stays off limits, so the run's own logs and state are never mistaken for task content.

The Dashboard opens in your browser automatically, and the console prints one line per role as the run progresses. At the end you get a plain-language reply that answers your request from the verified state alone, and says so plainly if the task did not finish.

Every run is stored under ./.lh-harness/runs/<run-id>/; the full report, including that reply, stays in the run's logs/report.json.

Verify the environment

lh-harness doctor

doctor is read-only. It reports the Python runtime, the agent CLIs, Node.js, and plugin state, and exits non-zero when a required check fails.

Agent CLIs are verified by running <binary> --version, not just by finding them on PATH, so one that is present but broken is reported as a failure instead of OK. This catches the Windows case where a Microsoft Store desktop install leaves a zero-byte codex.exe alias on PATH that is not the CLI; doctor prints how to fix it.

It also checks PyPI for a newer version. To check on its own:

lh-harness check-update

Configuration reference

lh-harness run reads ./.lh-harness/config.toml automatically. Precedence is:

Explicit CLI arguments
Values in ./.lh-harness/config.toml
Built-in defaults

Task text, run IDs, and API keys are deliberately not configurable here; they stay command-line or environment inputs so they never land in a file you might commit.

`[run]`

Field	Default	Description
`agent`	`"codex"`	Backend for every role unless a role overrides it: `codex` or `claude_code`.
`model`	`"gpt-5.6-sol"`	Model for every role unless a role overrides it. Must be a model the chosen backend exposes.
`env`	`"local"`	Execution environment. Only `local` today.
`runs_root`	`"./.lh-harness/runs"`	Where run directories are created. Each run gets `<runs_root>/<run-id>/`.
`workspace`	commented out	Working directory the agents operate in. Defaults to the directory `lh-harness` was started from, so a task acts on your real project; set it to isolate the run somewhere else.
`harness_dir`	commented out	Where harness task state is written. Defaults to the run's own `harness/`, keeping it out of the workspace.
`log_dir`	commented out	Where logs are written. Defaults to the run's own `logs/`.
`base_url`	commented out	OpenAI-compatible endpoint override, for a proxy or a self-hosted model.
`prompt_language`	`"en"`	Language of the harness-generated prompts and reports: `en` or `zh`. Does not restrict the task language.
`claude_mcp_config`	commented out	Path to a `.mcp.json` for Claude Code. Overrides the installed plugin.
`codex_mcp_config`	commented out	Path to a `[mcp_servers.*]` TOML for Codex. Overrides the installed plugin.
`mcp_add_dirs`	`[]`	Extra directories the MCP server may read. Claude Code rejects these, because its role isolation requires task files to live inside the workspace.
`max_rounds`	`30`	Upper bound on Manage-Execute-Audit rounds before the run stops.
`dashboard`	`true`	Start the web dashboard with each run.
`dashboard_port`	`0`	Dashboard port; `0` lets the OS pick a free one.

`[run.timeouts]`

Per-episode limits in seconds. One episode is a single role invocation, not the whole run.

Field	Default	Description
`manager`	`600`	Planning the next step.
`gui_executor`	`1800`	Executing a GUI/visual subtask.
`cli_executor`	`1800`	Executing a CLI/non-GUI subtask.
`auditor`	`600`	Verifying a subtask. Applies to both auditors.

`[run.roles.*]`

Each role can take its own agent and model, so you can pay for a strong model only where it matters: a capable Manager and Auditor with a cheaper Executor, for example. Every field is commented out by default, meaning "inherit".

Resolution walks the chain until it finds a value:

gui_executor → executor → [run].agent / [run].model
cli_auditor  → auditor  → [run].agent / [run].model

Section	Falls back to	Covers
`[run.roles.manager]`	`[run]`	The scheduler role
`[run.roles.executor]`	`[run]`	Both executor roles
`[run.roles.gui_executor]`	`executor`	GUI/visual subtasks
`[run.roles.cli_executor]`	`executor`	CLI/non-GUI subtasks
`[run.roles.auditor]`	`[run]`	Both auditor roles
`[run.roles.gui_auditor]`	`auditor`	GUI audit
`[run.roles.cli_auditor]`	`auditor`	CLI audit
`[run.roles.final_response]`	`manager`	The closing reply written for you

Every field above also has a CLI flag (--agent, --max-rounds, --gui-executor-model, --auditor-timeout, and so on) that overrides it for a single run. Run lh-harness run --help for the full list.

Manage computer-use plugins

Computer-use setup is intentionally separate from task execution: doctor only reports status, and lh-harness run never installs, removes, or changes plugins. All changes go through lh-harness plugin.

List the available plugins with their install state, supported agents, and homepages:

lh-harness plugin list

Plugin	Source	Agents	Platforms
`codex-computer-use`	Official plugin bundled with the Codex CLI	`codex`	whatever your Codex build offers
`open-computer-use`	npm (open-codex-computer-use)	`codex`, `claude_code`	macOS, Windows, Linux
`clawdcursor`	npm (clawdcursor)	`codex`, `claude_code`	macOS, Windows, Linux

Installing needs no agent flag. Every agent the plugin supports is configured, since the per-agent difference is only one more config file:

lh-harness plugin install clawdcursor

One install covers every project on the machine. It installs the package, runs whatever consent or permission step the plugin needs on the current OS, and writes one MCP config per agent under ~/.lh-harness/plugins/. Agents missing from PATH are skipped; --agent narrows the selection, and --no-activate skips the permission step on a headless machine.

lh-harness run then loads the right server automatically. When several are installed, the first available one wins:

codex-computer-use > open-computer-use > clawdcursor

--claude-mcp-config and --codex-mcp-config override that choice. plugin list and doctor both print which plugin each agent will load and whether its permissions are granted.

To remove one:

lh-harness plugin uninstall clawdcursor

GUI access stays scoped to the harness. The npm plugins live entirely inside ~/.lh-harness/ and are passed per run, so ~/.codex/config.toml, ~/.claude.json, and the user-scope MCP registries are never touched. codex-computer-use is the unavoidable exception: Codex loads it from its own registry, so codex plugin add records it there.

codex-computer-use needs manual grants on macOS. It raises no permission dialog, so an unauthorized GUI call just fails. The install opens the two panes for you; tick Codex Computer Use under Privacy & Security → Accessibility and → Screen & System Audio Recording, then re-run the install to verify. On Windows there is nothing to grant, but the harness has to run in a signed-in desktop session and stay unelevated.

Any missing prerequisite is printed during install.

Configure MCP servers

Any MCP server can be passed to the agents, not just computer-use ones. Each backend reads its own native format; nothing is translated between them.

Claude Code takes a .mcp.json file through --claude-mcp-config:

{
  "mcpServers": {
    "computer-use": {
      "command": "/path/to/mcp-server",
      "args": ["--option", "value"],
      "env": {
        "EXAMPLE_VARIABLE": "value"
      }
    }
  }
}

Codex takes a TOML file of [mcp_servers.<name>] tables through --codex-mcp-config, matching ~/.codex/config.toml:

[mcp_servers.my-server]
command = "/path/to/mcp-server"
args = ["--option", "value"]

[mcp_servers.my-server.env]
EXAMPLE_VARIABLE = "value"

Pass the config for the backend in use, plus any directory the server needs to read:

lh-harness run --task @task.md --agent codex \
  --codex-mcp-config /path/to/mcp.toml \
  --mcp-add-dir /path/to/mcp/files

Both flags can be given together when roles use different backends, and --mcp-add-dir may be repeated. The equivalent environment variables are LH_HARNESS_CLAUDECODE_MCP_CONFIG, LH_HARNESS_CODEX_MCP_CONFIG, and LH_HARNESS_MCP_ADD_DIRS, the last separated by : on macOS/Linux and ; on Windows.

Prefer letting the server read API keys from its environment over writing them into the config file.

Dashboard commands

lh-harness run --task @task.md --dashboard      # Monitor a live run
lh-harness dashboard                            # Browse completed and active runs

Common CLI options

Option	Description
`--task`	Task text or `@task.md`
`--agent`	`claude_code` or `codex`
`--env`	`local`
`--max-rounds`	Maximum number of Manage-Execute-Audit rounds; the CLI default is 30
`--dashboard`	Start live monitoring and human intervention
`--no-dashboard`	Disable a Dashboard enabled by the project configuration

Run a longer task from a file and open the Dashboard:

lh-harness run --task @task.md --dashboard

The Dashboard shows every round's plan, execution result, audit evidence, and reason for rework. It also provides human gates when a task completes, becomes blocked, needs input, or fails repeatedly.

📋 Plan	⚡ Execution	🔍 Audit	♻️ Rework
What happens next	What the agent did	What the environment proves	Why another round is needed

Every run is stored in an isolated runs/<run-id>/ directory. The complete task state and audit trail make the agent's progress inspectable, recoverable, and reproducible.

Run record	What it preserves
📋 Task state	Original goal, requirements, verified progress, and remaining work
🧾 Event stream	What happened throughout the run
🔍 Audit reports	Evidence and acceptance decisions for every round
🧠 Role trajectories	Manager, Executor, and Auditor inputs and outputs
📁 Workspace	Files and artifacts produced during execution
✅ Final report	The verified outcome of the task

Evaluation Reproduction

eval/ provides frozen reproduction suites for two benchmarks:

Directory	Benchmark	Description
`eval/WeaveBench-harness/`	WeaveBench (114 tasks)	Hybrid GUI+CLI tasks and a reproduction skill
`eval/OSWorldv2-harness/`	OSWorld-V2 (108 tasks)	Hybrid runner aligned with the official release

See each directory's README.md or README.zh-CN.md for environment setup, parameters, and launch commands. The nested cua_harness packages are frozen compatibility copies used for evaluation; new integrations should use src/lh_harness/.

Citation

@article{longhorizonharness2026,
  title={LongHorizon-Harness: Advancing Long-Horizon Agents for Real-World Tasks},
  author={Ziyu Ma and Hailang Huang and Shun Zou and Yong Wang and Shidong Yang and Yiming Hu and Fei Wei and XiangXiang Chu},
  journal={arXiv preprint arXiv:2608.01964},
  year   = {2026},
  url    = {https://arxiv.org/abs/2608.01964}
}

Operate the whole computer. Preserve verified progress. Keep working until the task is done.

File tree

1367

AMAP-ML/LongHorizon-HarnessView on GitHub

.github/
  workflows/
    release.yml
.gitignore
assets/
  harness_perf.png
  mea_main.png
  promotional_video_1440p.mp4
  quickstart.gif
  wechat_group.JPG
eval/
  OSWorldv2-harness/
    .gitignore
    cua-harness/
      src/
        cua_harness/
          __init__.py
          adapters/
            __init__.py
            base.py
            claude_code.py
            cli_agent.py
            codex.py
            openclaw.py
          cli.py
          environment/
            __init__.py
            base.py
            docker.py
            local.py
            ssh.py
          integrations/
            __init__.py
            vm.py
          llm.py
          orchestrator.py
          remote_io.py
          role_prompts.py
          runtime_signals.py
          storage.py
          time_utils.py
          types.py
          verifier_agent.py
    docs/
      EXPERIMENT_PARAMETERS.zh-CN.md
      OSWORLD_V2_LOCAL_CHANGES.zh-CN.md
      osworld-v2-vm-runtime.patch
    env.example
    OSWorld-V2/
      .claude/
        skills/
          analyze-task/
            prompts/
              analyze-single-task.md
            SKILL.md
            task-check-report.schema.json
          analyze-traj/
            prompts/
              analyze-full-run.md
              analyze-single-traj.md
            SKILL.md
          migrate-osworld-agent/
            SKILL.md
          setup-osworld/
            references/
              common.md
              cua-harness.md
              gitlab.md
              provider-aws.md
              provider-docker.md
              proxy.md
              unsupported-providers.md
              verify.md
              website.md
            SKILL.md
      .codex/
        skills/
          analyze-task/
            prompts/
              analyze-single-task.md
            SKILL.md
            task-check-report.schema.json
          analyze-traj/
            prompts/
              analyze-full-run.md
              analyze-single-traj.md
            SKILL.md
          migrate-osworld-agent/
            SKILL.md
          setup-osworld/
            agents/
              openai.yaml
            references/
              common.md
              cua-harness.md
              gitlab.md
              provider-aws.md
              provider-docker.md
              proxy.md
              unsupported-providers.md
              verify.md
              website.md
            SKILL.md
      .envrc
      .gitmodules
      .mise.toml
      .python-version
      assets/
        authorization.png
        aws-1.png
        aws-2.png
        aws-3.png
        banner.png
        creategcp.png
        desktopapp.png
        developer.png
        enableapi.png
        googleidentity.png
        googlephonecode.png
        googleshutoff.png
        netsetting1.png
        netsetting2.png
        netsetting3.png
        netsetting4.png
        oauth2.0.png
        oauthapp.png
        proxysetup-zh.png
        proxysetup.png
        pubeval_gdrive_auth.jpg
        pubeval_monitor1.jpg
        pubeval_monitor2.jpg
        pubeval_subnet.png
        pubeval1.png
        pubeval2.png
        pubeval3.png
        pubeval4.png
        pubeval5.png
        publishapp.png
        testusers.png
        unsafemode.png
        usertype.png
        vmware01.png
        vmware02.png
        vmware03.png
        vmware04.png
        vmware05.png
        vmware06.png
        vmware07.png
        vmware08.png
        vmware09.png
        winnetsetting1.png
        winnetsetting2.png
        winnetsetting3.png
        winnetsetting4.png
      benchmark_releases/
        osworld-v2-2026.06.24.json
        README.md
      desktop_env/
        __init__.py
        actions.py
        controllers/
          __init__.py
          gitlab.py
          moodle.py
          python.py
          setup.py
          website.py
          wine_game.py
        desktop_env.py
        evaluators/
          __init__.py
          backends/
            __init__.py
            anthropic_backend.py
            base.py
            bedrock_backend.py
            gemini_backend.py
            openai_backend.py
          getters/
            __init__.py
            calc.py
            chrome.py
            file.py
            general.py
            gimp.py
            impress.py
            info.py
            misc.py
            replay.py
            state.py
            vlc.py
            vscode.py
          metrics/
            __init__.py
            basic_os.py
            blender.py
            chrome.py
            directory_structure.py
            docs.py
            general.py
            gimp.py
            ics.py
            lab.py
            libreoffice.py
            llm_metrics.py
            obsidian.py
            others.py
            pdf.py
            slack.py
            slides.py
            survey.py
            table.py
            thunderbird.py
            utils.py
            vlc.py
            vscode.py
          model_client.py
          README.md
        file_source.py
        image_utils.py
        providers/
          __init__.py
          aliyun/
            __init__.py
            ALIYUN_GUIDELINE_CN.md
            ALIYUN_GUIDELINE.md
            config.py
            manager.py
            provider.py
          aws/
            __init__.py
            AWS_GUIDELINE.md
            config.py
            manager.py
            provider.py
            proxy_pool.py
            scheduler_utils.py
          azure/
            __init__.py
            manager.py
            provider.py
          base.py
          docker/
            DOCKER_GUIDELINE.md
            manager.py
            provider.py
          gcp/
            __init__.py
            manager.py
            provider.py
          README.md
          virtualbox/
            __init__.py
            INSTALL_VITUALBOX.md
            manager.py
            provider.py
          vmware/
            __init__.py
            disk.py
            INSTALL_VMWARE.md
            manager.py
            provider.py
          volcengine/
            __init__.py
            manager.py
            provider.py
            VOLCENGINE_GUIDELINE_CN.md
            VOLCENGINE_GUIDELINE.md
          volume.py
        safety/
          __init__.py
          disk.py
          escalation_guards.py
          files.py
          monitor.py
          system_config.py
        task_base.py
        user_simulator.py
      docs/
        EVALUATE_RESULT_JSON.md
        MIGRATING_FROM_OSWORLD_V1.md
        OSWORLD_SETUP_GUIDELINE.md
        PROVIDER_SETUP.md
        PROXY_GUIDELINE.md
        PUBLIC_EVALUATION_GUIDELINE.md
      evaluation_examples/
        __init__.py
        conflict_disambiguation.json
        cross_source_reasoning.json
        dynamic_environment.json
        examples/
          chrome/
            030eeff7-b492-4218-b312-701ec99ee0cc.json
            06fe7178-4491-4589-810f-2e2bc9502122.json
            0d8b7de3-e8de-4d86-b9fd-dd2dce58a217.json
            12086550-11c0-466b-b367-1d9e75b3910e.json
            121ba48f-9e17-48ce-9bc6-a4fb17a7ebba.json
            1704f00f-79e6-43a7-961b-cedd3724d5fd.json
            2888b4e6-5b47-4b57-8bf5-c73827890774.json
            2ad9387a-65d8-4e33-ad5b-7580065a27ca.json
            2ae9ba84-3a0d-4d4c-8338-3a1478dc5fe3.json
            3299584d-8f11-4457-bf4c-ce98f7600250.json
            35253b65-1c19-4304-8aa4-6884b8218fc0.json
            368d9ba4-203c-40c1-9fa3-da2f1430ce63.json
            3720f614-37fd-4d04-8a6b-76f54f8c222d.json
            44ee5668-ecd5-4366-a6ce-c1c9b8d4e938.json
            47543840-672a-467d-80df-8f7c3b9788c9.json
            480bcfea-d68f-4aaa-a0a9-2589ef319381.json
            59155008-fe71-45ec-8a8f-dc35497b6aa8.json
            6766f2b8-8a72-417f-a9e5-56fcaa735837.json
            6c4c23a1-42a4-43cc-9db1-2f86ff3738cc.json
            7a5a7856-f1b6-42a4-ade9-1ca81ca0f263.json
            7b6c7e24-c58a-49fc-a5bb-d57b80e5b4c3.json
            7f52cab9-535c-4835-ac8c-391ee64dc930.json
            82279c77-8fc6-46f6-9622-3ba96f61b477.json
            82bc8d6a-36eb-4d2d-8801-ef714fb1e55a.json
            93eabf48-6a27-4cb6-b963-7d5fe1e0d3a9.json
            9656a811-9b5b-4ddf-99c7-5117bcef0626.json
            99146c54-4f37-4ab8-9327-5f3291665e1e.json
            9f3f70fc-5afc-4958-a7b7-3bb4fcb01805.json
            9f935cce-0a9f-435f-8007-817732bfc0a5.json
            a728a36e-8bf1-4bb6-9a03-ef039a5233f0.json
            a96b564e-dbe9-42c3-9ccf-b4498073438a.json
            ae78f875-5b98-4907-bbb5-9c737fc68c03.json
            af630914-714e-4a24-a7bb-f9af687d3b91.json
            b070486d-e161-459b-aa2b-ef442d973b92.json
            b4f95342-463e-4179-8c3f-193cd7241fb2.json
            b7895e80-f4d1-4648-bee0-4eb45a6f1fa8.json
            bb5e4c0d-f964-439c-97b6-bdb9747de3f4.json
            c1fa57f3-c3db-4596-8f09-020701085416.json
            cabb3bae-cccb-41bd-9f5d-0f3a9fecd825.json
            da46d875-6b82-4681-9284-653b0c7ae241.json
            e1e75309-3ddb-4d09-92ec-de869c928143.json
            f0b971a1-6831-4b9b-a50e-22a6e47f45ba.json
            f3b19d1e-2d48-44e9-b4e1-defcae1a0197.json
            f5d96daf-83a8-4c86-9686-bada31fc66ab.json
            f79439ad-3ee8-4f99-a518-0eb60e5652b0.json
            fc6d8143-9452-4171-9459-7f515143419a.json
          gimp/
            045bf3ff-9077-4b86-b483-a1040a949cff.json
            06ca5602-62ca-47f6-ad4f-da151cde54cc.json
            2a729ded-3296-423d-aec4-7dd55ed5fbb3.json
            2e6f678f-472d-4c55-99cc-8e7c5c402a71.json
            38f48d40-764e-4e77-a7cf-51dfce880291.json
            554785e9-4523-4e7a-b8e1-8016f565f56a.json
            58d3eeeb-e9d0-499f-962e-fd0db2a744d8.json
            5ca86c6f-f317-49d8-b6a7-b527541caae8.json
            62f7fd55-0687-4a43-b6e1-3eda16fc6252.json
            72f83cdc-bf76-4531-9a1b-eb893a13f8aa.json
            734d6579-c07d-47a8-9ae2-13339795476b.json
            7767eef2-56a3-4cea-8c9f-48c070c7d65b.json
            77b8ab4d-994f-43ac-8930-8ca087d7c4b4.json
            7a4deb26-d57d-4ea9-9a73-630f66a7b568.json
            7b7617bd-57cc-468e-9c91-40c4ec2bcb3d.json
            8ea73f6f-9689-42ad-8c60-195bbf06a7ba.json
            a746add2-cab0-4740-ac36-c3769d9bfb46.json
            b148e375-fe0b-4bec-90e7-38632b0d73c2.json
            d16c99dc-2a1e-46f2-b350-d97c86c85c15.json
            d52d6308-ec58-42b7-a2c9-de80e4837b2b.json
            dbbf4b99-2253-4b10-9274-45f246af2466.json
            e19bd559-633b-4b02-940f-d946248f088e.json
            e2dd0213-26db-4349-abe5-d5667bfd725c.json
            f4aec372-4fb0-4df5-a52b-79e0e2a5d6ce.json
            f723c744-e62c-4ae6-98d1-750d3cd7d79d.json
            fbb548ca-c2a6-4601-9204-e39a2efc507b.json
          libreoffice_calc/
            01b269ae-2111-4a07-81fd-3fcd711993b0.json
            0326d92d-d218-48a8-9ca1-981cd6d064c7.json
            035f41ba-6653-43ab-aa63-c86d449d62e5.json
            04d9aeaf-7bed-4024-bedb-e10e6f00eb7f.json
            0a2e43bf-b26c-4631-a966-af9dfa12c9e5.json
            0bf05a7d-b28b-44d2-955a-50b41e24012a.json
            0cecd4f3-74de-457b-ba94-29ad6b5dafb6.json
            12382c62-0cd1-4bf2-bdc8-1d20bf9b2371.json
            1273e544-688f-496b-8d89-3e0f40aa0606.json
            1334ca3e-f9e3-4db8-9ca7-b4c653be7d17.json
            1954cced-e748-45c4-9c26-9855b97fbc5e.json
            1d17d234-e39d-4ed7-b46f-4417922a4e7c.json
            1de60575-bb6e-4c3d-9e6a-2fa699f9f197.json
            1e8df695-bd1b-45b3-b557-e7d599cf7597.json
            21ab7b40-77c2-4ae6-8321-e00d3a086c73.json
            21df9241-f8d7-4509-b7f1-37e501a823f7.json
            26a8440e-c166-4c50-aef4-bfb77314b46b.json
            2bd59342-0664-4ccb-ba87-79379096cc08.json
            30e3e107-1cfb-46ee-a755-2cd080d7ba6a.json
            347ef137-7eeb-4c80-a3bb-0951f26a8aff.json
            357ef137-7eeb-4c80-a3bb-0951f26a8aff.json
            37608790-6147-45d0-9f20-1137bb35703d.json
            3a7c8185-25c1-4941-bd7b-96e823c9f21f.json
            3aaa4e37-dc91-482e-99af-132a612d40f3.json
            4172ea6e-6b77-4edb-a9cc-c0014bd1603b.json
            4188d3a4-077d-46b7-9c86-23e1a036f6c1.json
            42e0a640-4f19-4b28-973d-729602b5a4a7.json
            4de54231-e4b5-49e3-b2ba-61a0bec721c0.json
            4e6fcf72-daf3-439f-a232-c434ce416af6.json
            4f07fbe9-70de-4927-a4d5-bb28bc12c52c.json
            51719eea-10bc-4246-a428-ac7c433dd4b3.json
            51b11269-2ca8-4b2a-9163-f21758420e78.json
            535364ea-05bd-46ea-9937-9f55c68507e8.json
            6054afcb-5bab-4702-90a0-b259b5d3217c.json
... 1208 more

FAQ

longhorizon-harness is a Claude Code plugin with 5 hand-picked skills for automation work, indexed on Flowy. Install it with the command on its page. It includes analyze-task, analyze-traj, migrate-osworld-agent. Its skills do not fire on their own yet. Request auto-invocation to have Flowy route them as you prompt. Free and open source.

What is longhorizon-harness?: The long-horizon computer-use harness. Run AI agents across desktop apps and the CLI for extended periods while preserving task state and making reliable progress on complex workflows. Features fresh-context execution, durable verified state, independent auditing, recoverable progress, and native Claude Code / Codex / OpenClaw integration.
How do I install longhorizon-harness?: Run these in Claude Code: npx -y skills add AMAP-ML/LongHorizon-Harness --agent claude-code. Then prompt normally.
Does longhorizon-harness auto-invoke its skills?: Not yet. It is indexed on Flowy as a plain plugin. Request auto-invocation on its page and Flowy will route its skills for you as you prompt.
Is longhorizon-harness free?: Yes. Flowy is free and open source, with nothing gated. You can read every skill in full before you install.