How to install WorkBuddy Skills and verify the first dry-run

To install WorkBuddy Skills from xplaai, start on the selected Skill page and use the WorkBuddy package when an Agent selector is provided. Ask WorkBuddy to import the archive, report the installed location and read the complete SKILL.md. Run the package's doctor check, configure the xplaai key through its supported local method, and stop after a dry-run until task count, models, output paths and paid actions are approved.

Browse xplaai Commerce Skills or create an API key.

Choose the package from the Skill page

xplaai's publishing standard uses separate Codex and WorkBuddy packages for verified dual-Agent releases. dog-video-auto-edit is one current example: its WorkBuddy package adds manager metadata used for recognition and maintenance. When the page offers separate packages, download the one labeled WorkBuddy. Do not rename a Codex archive and assume that it has gained the required format.

Package shape remains specific to each Skill. The free Product Video release currently has one shared xpla-video-gen ZIP instead of two Agent-specific files. Read the selected page and archive; do not turn either pattern into a universal rule.

Public case pages can explain a workflow, but a member package or access-controlled download remains protected. This guide does not expose private object paths or signed download links.

Import the ZIP into WorkBuddy

For a package designed for WorkBuddy, use Skills management to import the custom Skill ZIP or its extracted Skill folder. Refresh the Skills list after import. If the Skill does not appear, fully quit and reopen WorkBuddy before changing files.

Ask WorkBuddy to:

1. identify the package as a WorkBuddy Skill;
2. install it in the correct local location;
3. report the installed path;
4. read the included instructions;
5. run the package’s documented environment check.

Package commands differ. The instructions inside the selected Skill are the source of truth.

Use a bounded installation instruction:

Install this WorkBuddy Skill from the provided local archive.
Read SKILL.md and the installation notes completely.
Report the exact Skill name, installed path, documented doctor command,
missing dependencies and the supported dry-run.
Only report whether XPLA_API_KEY is configured; never show its value.
Do not call a paid API.

Verified WorkBuddy paths and commands for one real Skill

The current dog-video-auto-edit package has been verified against WorkBuddy Desktop 5.3.12. Its installation notes document these two targets:

From the extracted package, its user-level installer is:

python3 scripts/install_skill.py --target workbuddy

For a project-only installation, that same example supports:

python3 scripts/install_skill.py \
  --target workbuddy-project \
  --project /absolute/path/to/project

Replace the example project path with a real absolute path. Do not assume that another third-party Skill has this installer or these targets. Check its own documentation first.

Run doctor before an API call

The doctor step checks whether the local runtime, required tools and expected files are available. It should not perform a paid generation task.

If doctor reports missing software, fix that dependency and run it again. Do not keep launching paid tasks to diagnose a local setup problem.

For the verified pet-video example:

python3 scripts/doctor.py

The current doctor reports Python version, FFmpeg and FFprobe availability, Codex and WorkBuddy install locations, key presence, and separate readiness states. It requires Python 3.9 or later. It can reuse a WorkBuddy-bundled imageio-ffmpeg binary when present, while a full FFmpeg installation provides richer media inspection. Those are facts about this Skill, not requirements for all WorkBuddy Skills.

Keep the API key local

Create the key in the xplaai console and configure it through the method documented by the package. Never:

• paste a production key into a public repository;
• put the key inside the ZIP;
• include it in a screenshot or support message;
• ask WorkBuddy to print it back;
• hard-code it into browser JavaScript.

Use a placeholder such as XPLA_API_KEY in examples and store the real value in a local environment or protected configuration.

The pet-video example includes an interactive credential helper:

python3 scripts/configure_credentials.py
python3 scripts/configure_credentials.py --status

On macOS it can use Keychain; on Linux it can use Secret Service when secret-tool is available; on Windows its current user-environment fallback is persistent but not encrypted. The status command prints only configured or missing. Other packages may use another safe configuration method.

Start with a dry-run

A dry-run should report what the Skill intends to do before paid APIs are called. For a media workflow, review the scene, image, voice and video plan. For research, review the market, category, keywords and output destination.

Approve the plan only after verifying:

• the correct product or source material;
• the expected number and type of outputs;
• the chosen model or workflow;
• where files will be saved;
• which step can incur cost.

Use this review prompt:

Run only the documented dry-run with the source files in this project.
Report inputs, models or services, speech/image/video/data task counts,
reusable files, output paths, concurrency, estimated xplaai usage,
manual review gates and the exact flag that would enable live calls.
Wait for my approval. Do not run live generation.

Dry-run success does not prove that a final asset is accurate or publishable. For the first live attempt, approve one small sample, inspect it, and expand only after product facts, character consistency, media parameters and file paths are correct.

What the real screenshot proves—and does not prove

The screenshot above proves that the current xplaai pet-video detail page exposes a Codex/WorkBuddy selector and marks WorkBuddy as verified. It does not prove:

• that the current account has membership access;
• that every Skill installs to the same directory;
• that the doctor will pass on every machine;
• that the first generation will succeed;
• that a generated video will receive views, orders or revenue.

A complete installation record is the legitimate package source, actual install path, doctor report, key-presence status, dry-run plan and one reviewed sample. None of those records should include the full key or a private download URL.

Use the correct guide for Codex

If your Agent is Codex, follow the separate Codex Skill installation guide. The safety sequence is similar, but package metadata, recognized folders and installer targets can differ.

You can test a concrete workflow with the English Affiliate Video Skill or the TikTok Shop product research workflow.

Developers building their own orchestration should start with the xplaai API overview instead.

Frequently asked questions

Can I use the Codex ZIP in WorkBuddy?

When the Skill page offers separate packages, use the WorkBuddy archive. If the Skill documents one shared package, follow that release. Never rename an archive to claim compatibility.

Where does a WorkBuddy Skill install?

Let WorkBuddy follow the archive’s current instructions and report the installed path. Do not guess a universal path.

What does the doctor check do?

It checks the local environment and required dependencies before the workflow calls paid services.

Where should I store the xplaai API key?

Use the local protected configuration documented by the package. Never store it inside the Skill archive or a public project.

Why should the first run be a dry-run?

It lets you review inputs, planned steps, output location and paid actions before execution.

Are all Skill downloads public?

No. Public case information and access-controlled packages are different. Use the legitimate page and account flow for the selected Skill.

Does WorkBuddy itself choose the image or video model?

Not necessarily. The Skill workflow and xplaai configuration choose the services. For example, the current pet-video Skill uses XPLA speech, GPT Image 2, Veo 3.1 Fast and local FFmpeg. Another Skill may use a different stack.

Does installing a free Skill make API use free?

No. Package access and API consumption are separate. Review the dry-run and current account information before a live call.

How should I update an existing Skill?

Use the package's documented update flow. The verified pet-video installer supports --force, backs up the previous installation outside the active Skills directory, and then replaces it. Do not overwrite a working install until the new archive and its source are verified.

Install one Skill and verify it

Choose a WorkBuddy workflow, install the matching package, run doctor and approve a dry-run before the first paid task.