Blog Image Creative Director
Generate and edit publication-ready blog visuals with Gemini-powered prompt orchestration.
The skill clearly discloses Gemini MCP use, API-key requirements, output paths, and external dependencies; setup defaults to a mode-0600 user configuration, project mode avoids literal keys, and the npm package is pinned. Points are deducted because generation confirmation, cost consent, and outbound data boundaries are not explicit; `--key` may expose credentials through process listings or logs; prompts and images are sent to an external service, while `npx` retains supply-chain and network-runtime risk.
The happy path, MCP-unavailable fallback, common errors, and up-to-three safety rephrase attempts are described, and the setup and validation scripts check several configuration conditions. Points are deducted because this is a static review with no execution of key paths; tool behavior depends on the external package version, model availability descriptions are not fully consistent, and some failure behavior and response formats remain runtime-dependent.
The audience, commands, input forms, image types, ratios, and output fields are reasonably clear, with both standalone and internal invocation supported. Points are deducted because non-fit boundaries, trigger precedence, and offline alternatives are under-specified; core functionality depends on Google Gemini, Google AI, and npm network services, with no mainland-China reachability or Chinese-language workflow guidance.
The SKILL.md uses layered quick reference, workflows, error handling, and on-demand references, and includes MIT licensing, version metadata, argument hints, examples, limitations, and setup scripts; the dependency is version-pinned. Points are deducted because the selected skill does not clearly identify maintenance ownership, a changelog, or an update path; model, pricing, and service-status claims in references lack locally verifiable support, and the removal operation lacks confirmation or recovery guidance.
The skill specifies a coherent flow from intent analysis through six-component prompt construction, ratio selection, generation or editing, and delivery metadata. Its output includes a path, crafted prompt, settings, alt text, and frontmatter, making the core blog-image task plausibly usable. The static calibration cap limits this to 7; points are deducted because generated results were not verified, so quality, text accuracy, path correctness, and actual MCP compatibility remain uncertain and manual review/post-processing may be required.
The files provide tool parameters, a pinned npm version, configuration validation, error mappings, and security-handling details, giving some auditability. The static calibration cap limits this to 5; points are deducted because there are no committed tests or real CI execution evidence covering this skill's key generation, editing, and fallback paths, while pricing, rankings, accuracy, and service-status claims are primarily documentary assertions.
- Before using Google AI and nanobanana MCP, confirm that prompts and images are sent to an external service and assess cost, regional reachability, and data sensitivity.
- Avoid passing real credentials through `--key`; prefer environment variables or private user configuration, and review the `npx` package provenance and pin.
- Generation, editing, and fallback paths were not executed; add key-path tests and verify the documented models, pricing, ratios, and output files before production use.
What does this skill do, and when should you use it?
Blog Image is the dedicated visual-content skill in the claude-blog suite. It uses Gemini through MCP to generate or edit hero images, inline illustrations, social cards, and Open Graph images. Before generation, it analyzes the intended use, style, dimensions, and mood, then builds a six-component reasoning brief and selects a blog-focused domain mode. It requires nanobanana-mcp and a Google AI API key; when called internally without MCP, it fails gracefully so the parent workflow can continue with stock imagery.
Handles requests such as “generate blog image,” “generate hero image,” “social card,” and “edit blog image”; analyzes image type, article topic, style, dimensions, constraints, and mood; selects among Editorial, Product, Landscape, UI/Web, Infographic, and Abstract modes; constructs a natural-language prompt from Subject, Action, Context, Composition, Lighting, and Style; sets the aspect ratio before calling Gemini MCP tools; generates, edits, or iteratively refines images; creates descriptive 10–125 character alt text; returns the image path, crafted prompt, settings, domain mode, frontmatter example, and refinement suggestions; and can resize or convert outputs with ImageMagick for blog delivery.
- A solo blogger needs a 16:9 hero image or OG card for a newly written article and can run `/blog image generate`.
- A marketing team is preparing a product review or comparison and wants a studio-style product visual using Product mode.
- A technical writer needs a clean interface illustration, diagram, or data-oriented visual and can use UI/Web or Infographic mode.
- An editor wants to remove a background, adjust warmth or brightness, add text, or crop an existing image for social sharing with `/blog image edit`.
- A blog-write or blog-rewrite workflow needs optional generated visuals while remaining able to continue with stock photos if the MCP server is unavailable.
What are this skill's strengths and limitations?
- Purpose-built for blog heroes, inline illustrations, OG images, and social cards.
- Uses a six-component reasoning brief instead of forwarding raw user text to the image API.
- Supports six domain modes, image editing, aspect-ratio control, and alt-text generation.
- Gracefully falls back during internal blog-write or blog-rewrite calls when MCP is unavailable.
- Requires nanobanana-mcp, Node.js 18+, npx, and a Google AI API key.
- Direct generation depends on the Gemini MCP setup; without it, the user must run the setup command first.
- The source provides no dedicated test count or independent quality evaluation for this sub-skill.
- Detailed model pricing, rate limits, and MCP parameter specifications are referenced but not included in the supplied material.
How do you install this skill?
This skill is distributed as part of the AgriciDaniel/claude-blog collection; the source does not document a standalone folder-install procedure. With Claude Code 1.0.33+, install the public collection using /plugin marketplace add AgriciDaniel/claude-blog, followed by /plugin install claude-blog@agricidaniel-claude-blog, then restart Claude Code. Configure image generation with /blog image setup; the setup requires Node.js 18+, npx, and a Google AI API key, and can write MCP configuration to user or project settings.
How do you use this skill?
Run /blog image generate <idea> to create an image, /blog image edit <path> <instructions> to modify an existing image, or /blog image setup to configure the MCP server. Example: /blog image generate a 16:9 editorial hero image for an article about AI citation analytics. If the request is vague, the skill asks one clarifying question about intended use and style.
How does this skill compare with similar options?
Compared with sending raw text directly to an image model, Blog Image adds blog-use analysis, domain-mode selection, six-component prompt construction, aspect-ratio setup, and alt-text delivery. Compared with a general image generator, it is more tightly oriented toward blog publishing dimensions, OG cards, and frontmatter metadata.