
Addy Osmani Agent Skills
Apply explicit specification, testing, review and release criteria to AI-assisted engineering work.
Browse alphabetically. Open a product to see its sources and limits.

Apply explicit specification, testing, review and release criteria to AI-assisted engineering work.

A development environment for running and reviewing coding agents.

Choose AI SDK for the streaming application layer, then budget model access and hosting independently.

AI pair programming in your terminal with git integration.

Local coding agents with your own model keys or local models.

Choose Apify when an AI workflow needs a ready-to-run web tool with a documented input schema and a reviewable usage budget.

Choose Phoenix for operating your own AI trace and experiment store, after checking the ELv2 service restriction.

An AI application builder combining research, product planning and software creation workflows.

Draft a business website with AI, then personalize it alongside built-in business tools.

Choose Baseten after deciding whether you want a managed model API or to own the resource footprint of a dedicated deployment.

A local-first IDE that runs and coordinates multiple coding agents.

Choose Beam after deciding whether the AI workload needs a persistent machine, a short sandbox or a serverless GPU endpoint.

A terminal coding agent with planning and parallel agent execution.

Chat-based AI app and website builder with cloud backend.

Connect production traces to evaluation datasets while separating data processing, scoring and model inference costs.

Choose Browserbase when the agent needs managed browser capacity and you can separately track reasoning, search and fetch costs.

Choose Browserless when your agent already has browser automation code and needs managed browser capacity with explicit time limits.

Choose Cerebrium when you can estimate CPU, memory and GPU runtime separately for a deployed AI service.

Agentic coding that plans, writes code, runs tests, and opens pull requests.

Open-source coding agent for the IDE, terminal, and desktop.

AI-assisted pull-request review, change explanation and review prioritization.

AI code editor and coding agent for building software.

Choose Daytona when an agent needs an isolated machine whose compute and retained storage can be budgeted independently.

Autonomous software engineer that plans, writes, tests, and ships code.

Generate technical diagrams from text and continue editing them in Eraser.

Define a debugging team in YAML and run it from your terminal with your chosen models.

Use DSPy when you can define a scoring function and examples before optimising a language-model pipeline.

Evaluate a unified API by the model bill and credit-purchase fee, with private-deployment requirements treated as a separate plan decision.

Use Droid to plan, edit and validate repository changes from a desktop app or CLI.

Use a Gemini-powered terminal agent with file, shell and web tools.

AI coding assistant and agents integrated across GitHub and editors.

Choose Gladia after checking both model language coverage and the default training setting.

Google's agentic development platform spanning a desktop workspace, IDE and command line.

Build a queryable code knowledge graph with explicit versus inferred relationships for coding assistants.

Work on code from a terminal with a coding agent.

Monitor and steer Claude Code or Codex sessions from a mobile or web client.

Compose retrieval, routing, generation and agent components into explicit Python pipelines.

Budget Hostinger AI Website Builder using the upfront term, renewal price and one-time AI credit gift rather than the promotional monthly equivalent alone.

An open-source coding agent available through editors, a CLI and cloud workflows.

An agentic development environment that turns requirements into specifications and code changes.

Kombai connects UI design, repository context and browser refinement in a frontend-focused workflow.

Choose LanceDB Embedded when retrieval should live with the application, and separate that decision from an enterprise service purchase.

Trace multi-step agents and compare evaluation runs while keeping seat fees separate from execution and model costs.

Choose LangWatch by events per workflow, rather than treating one customer conversation as one event.

Choose a persistent-agent workspace by agent count, memory size and model usage, with personal and developer plans kept distinct.

Package and run a local language model as a portable executable instead of provisioning a model service.

Connect data sources and model providers to build indexing, retrieval and agent applications in Python.

Chat-driven AI software engineer that builds and hosts web apps.

Start Milvus with a local-file retrieval prototype, then choose production deployment and operations separately.

Choose Modal when you need to run your own AI code with explicit resource and execution control.

Assign coding-agent work to issues and keep progress and review in a shared board.

An office-style harness that delegates work across coding agents.

Assess NSFWJS for browser-side image classification, including TensorFlow.js setup, model delivery and the need to calibrate review thresholds on your own content.

Use a configurable terminal coding agent with language-server and debugger tools.

Cloud and CLI coding agent available across ChatGPT, editor, and terminal.

Work on a code repository from a terminal, desktop app or IDE, with your choice of model.

Operate coding agents across local, sandboxed or remote backends from Agent Canvas.

Run coding agents in separate Git worktrees and review their results together.

Control existing coding agents across desktop, mobile, web and CLI clients.

Open-source coding agent for terminal-based work with local files.

An AI website builder for service businesses with page editing and lead workflows.
A design-agent crew that turns a brief into HTML and Tailwind.

Guide a coding agent to reuse existing components and review changes for unnecessary complexity.

Compare gateway reliability controls and log retention with model costs kept outside the platform subscription.

Choose promptfoo for repeatable test configurations whose model combinations and assertions you can inspect.

Keep prompt revisions and evaluation datasets together when several people need to compare changes before rollout.

Validate Pydantic AI’s tool and output wiring offline before paying for an end-to-end model run.

Budget Qoder coding tasks with monthly credits, expiring top-ups and a reviewable repository workflow rather than treating credits as a fixed request count.

A website planning and building system using AI-generated sitemaps, wireframes and reusable components.

Prompt-to-app builder with built-in hosting, database, and authentication.

Separate app builders from internal users when costing an AI-enabled operational tool, then check the required governance tier.

Turn a screenshot or screen recording into an editable front-end prototype.

Select smolagents when code actions make a tool workflow easier to express, and isolate execution independently.

Evaluate Sourcegraph for cross-repository understanding and governance, using its current enterprise contract rather than a legacy Cody subscription.

Compare Speechmatics models by language handling and required features before comparing their hourly rates.

A conversational builder for business websites, applications and request-handling workflows.

Structure agent coding work around a reviewed design, implementation plan, tests and code review.

Let a configured language model use repository tools to propose issue fixes.

Control configured coding agents from desktop, mobile or web interfaces.

Serve AI code completion and repository-aware assistance from infrastructure you operate.

Tavus suits builders who need a real-time face-to-face AI interface and must budget simultaneous sessions separately from total usage.

Run coding agents in parallel worktrees from one development workspace.

Prompt-to-app builder that generates and deploys working web applications.

An AI development workspace that plans, implements and tests software from a project brief.

Use a terminal-based environment with a built-in coding agent or external CLIs.

Separate Weaviate database capacity from embeddings and Query Agent costs when budgeting retrieval.

Agentic editor for managing fleets of local and cloud coding agents.

Edit code collaboratively and run configured agents from a native editor.

Choose Zencoder’s team tier for shared credits and governance, rather than assuming Pro includes those controls.
Try another filter. Use the heart on a product to save it to your account.
Tell us what you need →Compare browser builders, editor assistants and terminal agents around the code you have and how you review changes.
Start a web app from a brief without installing a local editor. Check deployment, account requirements and export before committing.
Before choosing Confirm code export and budget for hosting, databases and connected services.
Work with an existing project inside an editor. Compare supported environments, project context and the way you review changes.
Before choosing Inspect the diff and run project checks; a generated change is not a verified fix.
Use a coding agent alongside a repository and command-line tools. Check installation, model access and action approvals.
Before choosing Review command permissions, generated patches and model usage costs.
89 listed products, by how you can start
Free access can still involve paid model usage, hosting or connected services. Check each product's pricing notes.
Choose an editor, terminal agent or app builder. Compare workflows and inspect a recorded bug fix.
Turn a brief into a working website or app, then review and publish it.
Compare 18 candidates →Give a coding agent a reproducible problem and review the changes it makes.
Compare 32 candidates →An IDE assistant fits work inside an existing editor. A terminal agent works directly with a repository. A browser builder helps start a new app. Choose the working environment before comparing model names.
For a new site, inspect source export, hosting and database dependencies. For an existing project, start on a branch, describe a reproducible issue and ask for a focused patch with validation.
There is no single winner across new apps, existing repositories and team environments. For an editor workflow, compare Cursor, GitHub Copilot and Cline. For terminal work, compare Claude Code, Codex, Aider and Pi. For browser-based app building, compare Lovable, Replit and Bolt. Run the same small task before committing to a paid plan.
Separate the client license, model inference and hosting. A free installation does not establish a free end-to-end task. Check the model provider before running a large repository.
These options let you choose model access; budget for provider usage or your own local compute.
Before a long run, separate three things: the cost shown by the client, the usage billed by the provider, and the rule that actually stops spending. A warning or a token counter does not establish a hard cap.
Cline shows an estimated task cost that updates after API requests. Reconcile it with the selected provider’s bill; it is not a guarantee of the final charge.
Aider’s /tokens command reports the current chat context. It is not a total of every billed request or retry in the task.
OpenCode Zen documents monthly usage limits separately from automatic balance reloads. An enabled reload can charge more than the usage limit; check both controls in your account.
Blank template only: no benchmark results, preset prices or product ranking. Fill it locally; this page does not upload the worksheet or its contents.
Official documentation reviewed on 2026-10-05. These are documentation checks, not hands-on product tests.
Give a coding agent a reproducible problem and review the changes it makes.
A repository, reproduction steps and expected behavior
Code changes and test results
Review the diff and rerun relevant tests. Generated code is not a security guarantee.
Choose based on repository access, local editor versus cloud execution, and how much control you need over command execution.

AI code editor and coding agent for building software.
Why it fits Cursor can work with an existing codebase to propose and review code changes for a reproducible problem.
PrepareAn accessible repository and reproduction steps
GetProposed code changes and an explanation
ReviewReview the diff and run your project tests

AI coding assistant and agents integrated across GitHub and editors.
Why it fits GitHub Copilot can work with an existing codebase to propose and review code changes for a reproducible problem.
PrepareAn accessible repository and reproduction steps
GetProposed code changes and an explanation
ReviewReview the diff and run your project tests

Agentic coding that plans, writes code, runs tests, and opens pull requests.
Why it fits Claude Code can work with an existing codebase to propose and review code changes for a reproducible problem.
PrepareAn accessible repository and reproduction steps
GetProposed code changes and an explanation
ReviewReview the diff and run your project tests

Cloud and CLI coding agent available across ChatGPT, editor, and terminal.
Why it fits OpenAI Codex can work with an existing codebase to propose and review code changes for a reproducible problem.
PrepareAn accessible repository and reproduction steps
GetProposed code changes and an explanation
ReviewReview the diff and run your project tests

Autonomous software engineer that plans, writes, tests, and ships code.
Why it fits Devin can work with an existing codebase to propose and review code changes for a reproducible problem.
PrepareAn accessible repository and reproduction steps
GetProposed code changes and an explanation
ReviewReview the diff and run your project tests

Agentic editor for managing fleets of local and cloud coding agents.
Why it fits Windsurf can work with an existing codebase to propose and review code changes for a reproducible problem.
PrepareAn accessible repository and reproduction steps
GetProposed code changes and an explanation
ReviewReview the diff and run your project tests

Open-source coding agent for the IDE, terminal, and desktop.
Why it fits Cline can work with an existing codebase to propose and review code changes for a reproducible problem.
PrepareAn accessible repository and reproduction steps
GetProposed code changes and an explanation
ReviewReview the diff and run your project tests

AI pair programming in your terminal with git integration.
Why it fits Aider can work with an existing codebase to propose and review code changes for a reproducible problem.
PrepareAn accessible repository and reproduction steps
GetProposed code changes and an explanation
ReviewReview the diff and run your project tests

Open-source coding agent for terminal-based work with local files.
Why it fits Pi can read a local workspace, run commands and write files as part of a developer-controlled coding workflow.
PrepareA local codebase and reproducible error
GetCode changes or a diagnostic explanation
ReviewInspect the diff and run the relevant tests

Work on code from a terminal with a coding agent.
Why it fits The CLI supports code work and terminal execution; our isolated bug-fix case preserves the exact generated patch and checks.
PrepareA code task or repository
GetProposed code changes
ReviewReview code changes

A local-first IDE that runs and coordinates multiple coding agents.
Why it fits bb launches multiple coding agents and lets one agent spawn and manage another.
PrepareA repository and a bounded development task
GetAgent threads, code changes and reviewable work
ReviewReview diffs and run project checks before merging

Run coding agents in parallel worktrees from one development workspace.
Why it fits Unstoppable runs coding agents in parallel Git worktrees with review tools.
PrepareA repository and separate development tasks
GetAgent changes in isolated worktrees and reviewable pull requests
ReviewReview plans, diffs and CI before integrating changes

An office-style harness that delegates work across coding agents.
Why it fits Its orchestrator assigns coding work to multiple CLI agents with a shared board and inbox.
PrepareA local project and a development goal
GetAssigned agent tasks, code changes and status
ReviewResolve questions, inspect code and run tests

A development environment for running and reviewing coding agents.
Why it fits ADE provides agent chats, subagents, worktrees and pull-request review in one environment.
PrepareA repository and an agent task
GetWorktree changes and pull requests
ReviewInspect diffs, checks and merge decisions

A terminal coding agent with planning and parallel agent execution.
Why it fits Blackbox CLI plans code changes and can dispatch several agents on separate branches.
PrepareA repository and a coding goal
GetPlans, code changes and reviewable diffs
ReviewApprove plans, inspect changes and run tests

Local coding agents with your own model keys or local models.
Why it fits The official page describes filesystem changes, model configuration and test feedback in a local coding-agent runtime.
PrepareA local codebase, change request and configured model
GetCode changes, diffs and test feedback
ReviewReview diffs and rerun the relevant tests before shipping

An agentic development environment that turns requirements into specifications and code changes.
Why it fits An agentic development environment that turns requirements into specifications and code changes.
PrepareA repository, reproduction steps and acceptance criteria
GetSpecifications and proposed code changes
ReviewReview the plan, diff and project test results

Google's agentic development platform spanning a desktop workspace, IDE and command line.
Why it fits Google's agentic development platform spanning a desktop workspace, IDE and command line.
PrepareA codebase, bounded issue and test commands
GetProposed code changes and development artifacts
ReviewReview tool permissions, diffs and verification results

Work on a code repository from a terminal, desktop app or IDE, with your choice of model.
Why it fits Use OpenCode to investigate a reproducible issue in a repository, then review the proposed patch.
PrepareA code repository and a reproducible task
GetProposed repository edits and explanations
ReviewChoose the model and check its billing and data policy

Use a Gemini-powered terminal agent with file, shell and web tools.
Why it fits Task links are limited to the documented workflow. Infrastructure and model components are not treated as ready-made end-user agents.
PrepareA project or task prompt and authorized Gemini access
GetFile changes, model responses and tool output
ReviewApprove shell access and verify code changes and external-source claims.

Monitor and steer Claude Code or Codex sessions from a mobile or web client.
Why it fits Task links are limited to the documented workflow. Infrastructure and model components are not treated as ready-made end-user agents.
PrepareAn authenticated coding CLI, a local project and a paired client
GetRemote session messages, approvals and underlying agent edits
ReviewVerify device pairing and review the underlying agent's commands and code diffs.

Assign coding-agent work to issues and keep progress and review in a shared board.
Why it fits Task links are limited to the documented workflow. Infrastructure and model components are not treated as ready-made end-user agents.
PrepareAn issue, a configured runtime and authenticated coding CLIs
GetIssue-linked run progress, agent comments and proposed code changes
ReviewReview the issue's run history and diff before approving its completion.

Use a configurable terminal coding agent with language-server and debugger tools.
Why it fits Task links are limited to the documented workflow. Infrastructure and model components are not treated as ready-made end-user agents.
PrepareA repository, coding task and configured model provider
GetCode modifications, diagnostics and debugger or command results
ReviewReview tool permissions and code diffs and rerun the relevant tests.

Operate coding agents across local, sandboxed or remote backends from Agent Canvas.
Why it fits Task links are limited to the documented workflow. Infrastructure and model components are not treated as ready-made end-user agents.
PrepareA repository, a task and an authorized agent backend
GetAgent conversation records and proposed repository changes
ReviewUse an isolated backend and inspect diffs and commands before approving changes.

Run a desktop AI coworker with governed tools and security-review workflows.
Why it fits Task links are limited to the documented workflow. Infrastructure and model components are not treated as ready-made end-user agents.
PrepareAn authorized repository or local task and configured model access
GetSecurity findings, proposed fixes and audited task results
ReviewValidate scanner findings, review remediation diffs and approve every consequential action.

Run coding agents in separate Git worktrees and review their results together.
Why it fits Task links are limited to the documented workflow. Infrastructure and model components are not treated as ready-made end-user agents.
PrepareA Git repository, configured coding CLIs and a task prompt
GetIsolated agent worktrees, diffs and selected code changes
ReviewCompare candidate diffs and tests before merging a worktree into the main project.

Control existing coding agents across desktop, mobile, web and CLI clients.
Why it fits Task links are limited to the documented workflow. Infrastructure and model components are not treated as ready-made end-user agents.
PrepareAn authenticated coding CLI, a local daemon and a paired client
GetRemote agent progress, approvals and underlying code changes
ReviewConfirm device authorization and review commands and diffs from the underlying agent.

Let a configured language model use repository tools to propose issue fixes.
Why it fits Task links are limited to the documented workflow. Infrastructure and model components are not treated as ready-made end-user agents.
PrepareA reproducible repository issue, agent configuration and model access
GetA proposed patch and execution trajectory
ReviewReview the trajectory and diff and rerun the failing tests before accepting a patch.

Control configured coding agents from desktop, mobile or web interfaces.
Why it fits Task links are limited to the documented workflow. Infrastructure and model components are not treated as ready-made end-user agents.
PrepareA supported authenticated coding agent and a local development project
GetAgent-session controls, approvals and underlying repository edits
ReviewCheck remote access and inspect commands and diffs before accepting agent work.

Use a terminal-based environment with a built-in coding agent or external CLIs.
Why it fits Task links are limited to the documented workflow. Infrastructure and model components are not treated as ready-made end-user agents.
PrepareA repository, development task and configured agent access
GetTerminal output, code modifications and agent task results
ReviewReview shell execution and code changes and confirm cloud-service billing separately.

Edit code collaboratively and run configured agents from a native editor.
Why it fits Task links are limited to the documented workflow. Infrastructure and model components are not treated as ready-made end-user agents.
PrepareA project, development request and configured agent or model access
GetEditor changes, agent tool output and reviewable diffs
ReviewReview code changes and tool permissions and rerun relevant project tests.

Define a debugging team in YAML and run it from your terminal with your chosen models.
Why it fits The stable documentation includes an investigator-and-fixer workflow for repository bugs.
PrepareA YAML agent configuration, a repository and a reproducible error
GetAgent responses, tool output and proposed code changes
ReviewChoose models, review their billing and set permitted tools before running.

Use Droid to plan, edit and validate repository changes from a desktop app or CLI.
Why it fits Use Droid to plan, edit and validate repository changes from a desktop app or CLI.
PrepareA repository, a reproducible issue and acceptance checks
GetProposed code changes, command output and validation results
ReviewReview every diff and rerun project checks before accepting changes.