CommunityCoding & Developmentgithub.com

creative-project

Design or revise branding, visual identities, logos, posters, illustrations, and social-media assets using project artwork or image references. Inspect the existing visual language, choose references, and iterate on named candidates.

What is creative-project?

creative-project is a Claude Code agent skill that design or revise branding, visual identities, logos, posters, illustrations, and social-media assets using project artwork or image references. Inspect the existing visual language, choose references, and iterate on named candidates.

Works with~Claude Code~Codex CLI~Cursor
npx skills add https://github.com/clawdbot/clawdbot/tree/main/skills/creative-project

Installed? Explore more Coding & Development skills: steipete/bluebubbles, steipete/eightctl, steipete/blucli · View all 6 →

Ask in your favorite AI

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

Documentation

Creative project

  1. Understand the request before choosing a look. Identify the deliverables, intended audience, exact copy, dimensions/crop requirements, and what the person wants preserved or changed. For an edit, resolve the named candidate or source attachment, not the newest image. Use available project context before asking questions. Make reasonable reversible assumptions; ask one focused question only when a missing decision would materially change the work. Check: the requested outcome and still-active constraints are clear.

  2. Inspect the actual visual references. For requests such as “use the designs in this repo,” locate relevant artwork, finished layouts, editable sources, and project guidance in the authorized workspace. Inspect representative full compositions as images, not just filenames, source code, isolated crops, or color samples. A poster's character grid, typography, hierarchy, and negative space can matter more than its pictured subject. Use a labeled contact sheet when comparing several sources helps; keep important details readable. Follow the active image-reading tool's schema. A file listing or image-loading acknowledgment is not an aesthetic assessment. Check: observations identify the actual reference compositions and their visual treatment.

  3. Choose a direction from those observations. Separate what the references show from your proposed interpretation: medium, typography, geometry, composition, texture, palette, and visual hierarchy. Preserve an existing design system unless the person requests a departure. Do not reinterpret every reference as support for a mascot, palette, slogan, or layout you already invented. Assign each selected input a purpose: style reference, composition reference, identity reference, edit target, or supporting content. A new design may use references without being an edit of their subject. Check: the proposed treatment follows the request and each reference has a reason to be used.

  4. Keep revisions and related outputs coherent. For a substantial multi-output or multi-turn project, maintain one concise named review document alongside the deliverables. Record the current brief, stable candidate names, the selected sources and their purposes, protected elements, rejected directions, latest requested changes, and the current output for each deliverable. Link existing files/attachments rather than copying the conversation or creating another media store. For a simple one-image request, the conversation can carry this context; do not create a project manifest unnecessarily. Reopen the current brief and relevant visuals on continuation or after context loss. Never infer human approval from recency, generation success, or delivery. Check: “revise B” and the current icon/header/poster resolve unambiguously.

  5. Use the appropriate creation path. Prefer editing existing SVG, HTML/CSS, canvas, or other native sources for deterministic typography, ASCII/character grids, precise layouts, and work extending those assets. Use the available image-generation/editing tool for requested raster illustration, photography, textures, or image transformations. Do not replace requested raster artwork with a low-fidelity code placeholder, or rasterize an editable design without a reason. Inspect unseen local edit targets first. Pass the selected image inputs explicitly using the tool's advertised reference mechanism; earlier inspection does not automatically forward them to a generator. Check supported provider/model capabilities and returned dimensions rather than assuming an option worked. Check: the chosen path preserves the requested medium and the actual request includes its intended references.

  6. Create and inspect the result. Follow the active tool's synchronous or background completion contract. When the generator supports privately held review candidates for this chat, choose that mode before generation, inspect the retained candidate through its advertised interface, then explicitly release only the current result for delivery. Otherwise follow automatic delivery without promising that inspection can withhold an attachment already sent. Do not restart a pending task because a person asks for progress. Inspect completed outputs as images against the current request and selected references: subject, medium, composition, typography and exact copy, protected elements, crop safety, and requested dimensions. Judge layout at its intended use size, not only zoomed in. State an unavailable visual inspection honestly; filenames, successful execution, and programmatic dimensions do not prove the design looks right. Check: each candidate has an observed result and any unmet requirement is identified.

  7. Revise deliberately. For a local correction, describe both the change and the elements that must stay unchanged. For a major style reset, release rejected style/composition references and conflicting self-imposed constraints, while preserving still-active copy and requirements. Inspect the revised result again. Use supported exact-task supersession or review controls when available; do not assume a new request cancels an older job, an abort retracts a delivered image, or “latest wins” applies to independent requests. Keep a late obsolete result distinct from the current design. Check: the next request addresses the feedback without silently reverting accepted work or reusing a rejected direction.

  8. Deliver usable work. Show the requested artifacts through the conversation's supported attachment/preview path, not merely a local path or a claim that they exist. Reuse host-delivered attachments rather than sending them twice. Label candidates and meaningful revisions concisely. Keep project-bound final assets and useful editable sources in the authorized project/workspace, not only temporary generation storage; preserve requested output formats and names. Update the review document when one exists, distinguishing current, rejected, superseded, and explicitly approved work. Do not publish to a public account or change sharing permissions without that separate authority. Check: the person can inspect the right result and the project can use the retained files.

Individual skills in this repo

This repo contains 20 individual skills — each has its own dedicated page.

1password

Set up and use 1Password CLI for sign-in, desktop integration, and reading or injecting secrets.

acp-router

Route plain-language requests for Claude Code, Cursor, Copilot, OpenClaw ACP, OpenCode, Gemini CLI, Grok Build, Qwen, Kiro, Kimi, iFlow, Factory Droid, Kilocode, or explicit ACP harness work into either OpenClaw ACP runtime sessions or direct acpx-driven sessions ("telephone game" flow). For coding-agent thread requests, read this skill first, then use only `sessions_spawn` for thread creation. Codex chat binding defaults to the native Codex app-server plugin unless ACP is explicit or background spawn needs ACP.

add-model-provider

Add and live-prove a model provider with non-interactive config one-liners, without exposing credentials.

agent-transcript

Requested GitHub PR/issue agent transcripts: redact, trim, preview, and insert safely.

apple-notes

Create, view, edit, delete, search, move, or export Apple Notes via the memo CLI on macOS.

apple-reminders

List, add, edit, complete, or delete Apple Reminders and reminder lists via remindctl.

auto-qa

Run OpenClaw-wide autonomous QA and live/stress campaigns across independent subsystem lanes, with verified fixes and a resumable evidence report.

autoreview

Structured code review through the shared OpenClaw agent-skills installation.

bear-notes

Create, search, and manage Bear notes via grizzly CLI.

blacksmith-testbox

Run Blacksmith Testbox for CI-parity checks, secrets, hosted services, migrations, or builds local cannot reproduce.

block-kit

Use proactively for structured or interactive Slack replies, and when asked to author or validate native Slack Block Kit JSON.

blogwatcher

Monitor blogs and RSS/Atom feeds for updates using the blogwatcher CLI.

blucli

BluOS CLI (blu) for discovery, playback, grouping, and volume.

bluebubbles

Send and manage iMessages via BlueBubbles, including attachments, tapbacks, edits, replies, and groups.

browser-automation

Use when controlling web pages with the OpenClaw browser tool, especially multi-step flows, login checks, tab management, or recovery from stale refs/timeouts.

camsnap

Capture frames or clips from RTSP/ONVIF cameras and local webcams, including USB pan/tilt/zoom control.

canvas

Present HTML on connected OpenClaw node canvases, navigate/eval/snapshot, and debug canvas host URLs.

channel-message-flows

Use when running QA Lab channel message flow evidence.

clawdtributor

Clawtributor PRs here, last week or another window: discover conversation refs, recheck GitHub, rank by impact.

clawhub

Search ClawHub for plugins by default, or skills and agent Templates when requested; add Templates; install, verify, update, uninstall, publish, or sync skills.

Related Skills