CommunityImagegithub.com

logo-prompting

Interviews the user about the existing brief and logo, then writes detailed, ready-to-paste AI image prompts for logo concepts or refinements. Activate whenever the user asks for a logo prompt or an earlier logo attempt came out wrong.

What is logo-prompting?

logo-prompting is a Claude Code agent skill that interviews the user about the existing brief and logo, then writes detailed, ready-to-paste AI image prompts for logo concepts or refinements. Activate whenever the user asks for a logo prompt or an earlier logo attempt came out wrong.

Works with~Claude Code~Codex CLI~Cursor
npx skills add https://github.com/mostafaibrahem847-sudo/Branding-skills/tree/HEAD/skills/logo-prompting

Installed? Explore more Image skills: steipete/songsee, affaan-m/frontend-design, mattpocock/improve-codebase-architecture · View all 6 →

Ask in your favorite AI

Open a new chat with this agent skill pre-loaded.

Documentation

Logo Prompting

Writes image-generation and revision prompts for logo concepts, then helps evaluate the outputs the user brings back. Prompt-only output is the default; do not assume this agent can generate image files. This skill requires an approved ../brand-identity brief. Use references/prompt-patterns.md for prompt structure and both shared reference libraries for precedent methods that must be translated rather than copied.

When to Use

  • User wants a first logo concept generated by AI
  • An AI-generated logo came back wrong (off-brand, wrong colors, unrelated shapes, bad text rendering) and needs a corrected prompt
  • User wants a small, specific change to an existing AI-generated logo (not a full redo)

Workflow

  1. Interview first: read ../shared-references/brand-intake-interview.md, inspect the brief and current logo assets, and ask only for missing or conflicting facts that affect the logo. Confirm exact spelling/scripts, market/audience, intended use, approved colors, existing logo status, and whether the user wants a new concept or a revision. Summarize the lock and wait for confirmation before writing a prompt.
  2. Choose one route from references/prompt-patterns.md (icon, wordmark, emblem, monogram, or an approved combination). Explain why the route fits. If the brief supports two routes equally, offer them as separate choices instead of mixing them in one prompt.
  3. Build a complete, copy-ready prompt with one concrete idea, shape logic, exact palette hex codes and roles, exact background hex, composition, intended-size test, tool/ratio when known, relevant exclusions, and a validation checklist. For bilingual or non-Latin names, copy exact user-approved text and specify manual typesetting if spelling is unreliable.
  4. Hand off and review: present the prompt and any reference assets to attach. The user runs it in their image tool and can return the result for review. Check brief fit, category distinctiveness, spelling, silhouette, one-color use, small-size reading, and similarity to references/competitors. If exact lettering or geometry is wrong, label the image as rough direction and recommend rebuilding the mark in an appropriate design tool.
  5. Iterate one decision at a time using references/iteration-strategy.md. Keep the actual approved logo attached when refining or applying it; record the prompt and rejected directions.
  6. Lock production assets only after user approval. Save approved files to /assets/logo/ only when the user asks for project-file handling (see ../brand-presentation/references/page-structure.md). Record actual formats, color variants, script/spelling, exact colors, measured/approved clear space, and known limits.

Hard Rules (from real failure cases — see common-pitfalls.md for the full list)

  • Never write "abstract shape" without anchoring it to something concrete the viewer will recognize (a basket, a crystal, a cup) — pure abstraction reliably produces unrelated, unusable shapes.
  • Never ask an AI image tool to render small, precise text (especially non-Latin scripts) inside a complex shape (e.g. letters forming a silhouette) without expecting several failed attempts — flag this risk to the user before they spend generations on it, and offer the fallback (icon + separate wordmark, not fused).
  • Always specify the exact hex codes, not "warm brown" or "a nice red" — color drift between generations is one of the most common complaints.
  • Always specify flat/vector/geometric style explicitly if that's the direction, or the tool defaults to literal, cluttered, or photorealistic renders that don't scale down as a usable mark.
  • When revising one part of an existing logo, describe the ENTIRE logo's current state in the prompt (what to keep, exactly) plus the one change — a prompt that only describes the change tends to regenerate the whole thing differently.
  • A generated raster image is a concept, not automatically an editable logo, vector file, trademark clearance, or accurate wordmark. Label it accurately and verify production needs separately.
  • Use references to study methods. Never request a “same as” recreation or reuse a distinctive competitor/reference device.
  • When the user requests a visual logo-guideline board, also read ../brand-guidelines-doc/references/visual-guidelines-prompt.md. Treat the board as a separate deliverable from the logo concept prompt; never invent clear-space or minimum-size rules.
  • Do not generate or edit images unless the user explicitly asks and an image-generation capability is available. A prompt request means return prompt text for the user's external tool.

Individual skills in this repo

This repo contains 1 individual skill — each has its own dedicated page.

Related Skills