Blog Image#

Use this skill to generate one blog thumbnail per run using the pixr CLI (npm package pixr-cli, version 0.3.0 or newer).

pixr works with several providers (Gemini, OpenAI-compatible, OpenRouter, fal) and has no built-in model. The provider, model, and API key come from the user's own pixr config. Never pick or change them yourself.

If the user did not provide a topic, ask for one before proceeding.

Inputs to infer#

• TOPIC: the subject of the blog post the image is for (e.g. "vector databases", "building a notes app in Rust")
• Optional: --save-to path override (default: ~/blog-images/)
• Optional: --prefix for the output filename

Prerequisite check#

Before generating, run these checks in order:

1. Confirm pixr is available:

bash

pixr --help

If this fails, run:

bash

npm install -g pixr-cli

Then verify again. If still failing, stop and report.

2. Confirm the API key and model are configured:

bash

source ~/.zshrc && pixr config --json

Read hasApiKey, provider, and model from the output.

• hasApiKey must be true. pixr reads PIXR_API_KEY (the older NANO_IMAGE_API_KEY also works), or the variable named by apiKeyEnv in the user's pixr config. These are defined in ~/.zshrc.
• model must not be null. pixr has no default model.

CRITICAL NON-NEGOTIABLE: never ask the user to add or set the API key manually. It is already in ~/.zshrc. Always source it before concluding the key is missing. If hasApiKey is still false after sourcing, report the exact output and stop.

If model is null, list the models for the user's provider and ask the user which one to use. Do not choose one yourself:

bash

source ~/.zshrc && pixr models --no-interactive

Save the model they pick with pixr model <name>, then run the config check again.

Workflow#

1. Confirm the topic with the user or infer it from the blog name they provided.
2. Create ~/blog-images/ if it does not exist.
3. Run the generation command using this exact structure (source ~/.zshrc in the same shell first):

bash

pixr generate -w 1600 -f webp --save-to ~/blog-images --no-stdin \
  "1920x1080 thumbnail on topic \"{TOPIC}\" dont just use text use proper vectors resarch on web and make and all, add more vectors then just text (less text more vectors images and all)"

Replace {TOPIC} with the actual topic. Keep the rest of the prompt exactly as written. Do not paraphrase, shorten, or reword the prompt text. It is battle-tested.

1. Confirm the output file was written to ~/blog-images/. pixr prints saved image: <absolute path>; note that exact path. Warnings are printed as warning: ...; report them but they do not fail the run.
2. Update the blog-meta JSON with the image path.
   • Look for a matching JSON file in ~/blog-meta/ whose blog_path value corresponds to the same blog.
   • If found, set thumbnail_path to the exact absolute path of the generated image file.
   • Write the updated JSON back to the same file. Do not change any other field.
   • If no matching JSON file is found, note it in the completion report but do not fail the run.
3. Report the full output file path.

Quality bar#

• CRITICAL NON-NEGOTIABLE: always use -w 1600 -f webp. Never change these flags.
• CRITICAL NON-NEGOTIABLE: always use --save-to ~/blog-images unless the user explicitly provides a different path.
• CRITICAL NON-NEGOTIABLE: never modify the prompt text. Use it verbatim with only {TOPIC} substituted.
• CRITICAL NON-NEGOTIABLE: do not skip the prerequisite check. If pixr is missing, install it first.
• CRITICAL NON-NEGOTIABLE: never choose a provider or model for the user. Use their pixr config; if no model is set, ask them.
• CRITICAL NON-NEGOTIABLE: after a successful generation, always attempt to update the matching ~/blog-meta/*.json with thumbnail_path. Never skip this step.

Completion report#

At the end, report:

• Topic used
• Full pixr command run
• Output file path
• Whether the file exists at that path after generation
• Which ~/blog-meta/*.json file was updated (or that no match was found)