How to use the KrillinAI Cover Generator agent skill

What the krillinai-cover skill makes the agent do

An agent skill is a folder of instructions that an AI agent loads when a task matches the skill's description. This one covers a single job: generating a cover image with the KrillinAI CLI from a complete image prompt.

The agent runs the KrillinAI cover command with your prompt, saves the result into a task directory, and then checks the output. It does not decide the title text, language, layout or style. Those must already be in the prompt you supply or that the agent writes from your instructions.

• Runs the KrillinAI cover command with a supplied prompt and a size (default 1024x1024).
• Writes generated_cover.png, cover_prompt.final.txt and krillinai_manifest.json.
• Confirms the image is non-empty and decodable, and that the saved prompt matches what was submitted.
• Reports errors rather than treating a manifest alone as success.

Install the krillinai-cover skill

Download the skill ZIP from this page and unzip it. For a personal install, place the krillinai-cover folder in ~/.claude/skills/. For a single project, place it in that project's .claude/skills/ folder.

In the Claude apps you can instead upload the ZIP under Settings → Capabilities → Skills.

• Personal: ~/.claude/skills/krillinai-cover/
• Project: .claude/skills/krillinai-cover/ inside your project
• Claude apps: Settings → Capabilities → Skills, then upload the ZIP

Set up KrillinAI before you install

The skill depends on the KrillinAI CLI and its configuration. Without them it will fail. The skill expects the CLI binary path in the KRILLINAI_CLI environment variable and an execution working directory in KRILLINAI_CWD.

The execution working directory must contain config/config.toml with an [image] provider section and an API key for an image-generation provider. The config file itself is not included with the skill, so you need to set it up in KrillinAI. The skill also reads skills/krillinai-cli/references/cli-contract.md for binary location, working directory and JSON behaviour, and that file is not included in the submitted files.

• KRILLINAI_CLI: path to the KrillinAI CLI binary
• KRILLINAI_CWD: KrillinAI execution working directory
• config/config.toml in that directory with an [image] provider section
• An image-generation provider API key set in config.toml

Trigger the skill with a natural request

You do not need to type a command. Ask the agent for a cover in plain language, and include the full prompt. The skill is triggered by requests to generate a cover image with the KrillinAI CLI from a complete image prompt.

A good request states the title text exactly, the language, the composition, the visual style and any useful context from the source video. Anything you leave out will not be added by the skill.

• Give the exact title text and language, for example the words to appear on the image.
• Describe composition: where the subject sits, where the text goes.
• Describe visual style and mood.
• Mention relevant source-video context if the image should reflect it.

Example: a complete prompt and the command it runs

The agent runs the cover command from the working directory, passing your workdir, the prompt and the size. This is the command pattern from the skill's instructions:

• --workdir: an existing or new task directory
• --prompt: the final prompt sent to the image model
• --size: optional, default 1024x1024; the skill's example uses 1536x1024

Check the output and the saved prompt

Do not stop at the first message saying the cover was made. The skill requires terminal JSON with "ok": true, then reads the output paths from the response or manifest. It confirms that generated_cover.png is non-empty and decodable as an image, and that cover_prompt.final.txt records exactly the prompt you submitted.

If the provider configuration is missing or generation fails, the agent should report the returned error. Treat any failure as a failure, not as a success because a manifest exists.

To test without spending a generation call, use a dry run. The skill notes that cover --dry-run validates the command and writes a manifest, but does not call the image model or create generated_cover.png.

• Success requires "ok": true in the terminal JSON.
• Check generated_cover.png is non-empty and a valid image.
• Check cover_prompt.final.txt matches your submitted prompt.
• Use --dry-run to validate the setup first.

Safety notes: scripts and network access

This skill runs scripts and uses the network. It executes the KrillinAI CLI binary and sends your prompt to the image-generation provider configured in config.toml. Review the prompt before generating, since it leaves your machine for that provider. Only install the skill if you trust the KrillinAI binary you point it at.

• Runs a local binary (KRILLINAI_CLI) and writes files to the task directory.
• Sends the prompt and generation request to your configured image provider over the network.
• Your provider API key lives in config.toml; keep that file private.

Tips and limits

The skill is only as good as the prompt. If a title or style matters, write it into the prompt explicitly. The CLI does not expose a reference-image argument and does not automatically send origin_cover.jpg to the model, so if you want the source cover reflected, describe it in the prompt instead.

• Write the title text, language, composition and style into the prompt.
• To mirror the source cover, describe it in words; the file is not sent as a reference.
• Dry-run first when changing configuration.
• Keep a task directory per cover so the outputs stay together.