🎨 Imagine

An art-direction engine for text-to-image generation that separates Style from Content to keep aesthetic identity consistent across an image series.

Skill packages follow the agentskills.io open standard and are compatible with OpenClaw, Hermes Agent, and any agentskills.io-compliant client.

---

Overview

Imagine treats image generation as a two-part process: a Style Prompt that fully describes aesthetic DNA (palette, lighting, composition, texture, framing) and a Content Prompt that describes only what is in the scene. Decoupling the two prevents style drift across a series and makes a captured style reusable as a named library entry.

The skill ships a curated library of baseline styles, extracts new styles from reference images via a multi-modal analysis prompt, and generates images through the Pollinations.ai Flux endpoint. Every run is journaled.

Commands

Command	Description
imagine.generate	Generate an image in a specific style.
imagine.extract	Analyze a reference image and produce a structured Style Prompt.
imagine.library.list	List all available predefined and saved styles.
imagine.style.save	Save a new custom style to the library.
imagine.journal	Read the most recent run record.

Setup

imagine.init runs on first use and creates {agent_root}/commons/data/ocas-imagine/ (with config.json, styles.jsonl, history.jsonl) and {agent_root}/commons/journals/ocas-imagine/. No manual setup is required. No API keys, accounts, or credentials are needed — Pollinations.ai is used anonymously.

Dependencies

External

• Pollinations.ai Flux endpoint (required) — zero-cost, no auth. See references/api_reference.md.
• Multi-modal LLM with image analysis (required for imagine.extract) — e.g. vision-capable Claude or GPT model.

Changelog

v1.0.4 — April 14, 2026

• Storage paths migrated from ~/openclaw/... to {agent_root}/commons/... per updated storage convention spec.
• All path references updated across SKILL.md, README.md, and references/journal.md.

v1.0.3 — April 14, 2026

• Split references/default_styles.md into one file per style (soma.md, noir.md, hiro.md, comic.md, candy.md, vaporware.md); default_styles.md is now an index.
• Support File Map in SKILL.md lists each style file individually.

v1.0.2 — April 14, 2026

• Moved storage from {agent_root}/commons/… to {agent_root}/commons/… per spec-ocas-storage-conventions.md.
• Added required System-skill sections (Responsibility Boundary, Optional Skill Cooperation, Ontology Mapping, Journal Outputs, Visibility, Background Tasks, Validation Rules).
• Added references/journal.md (record schema).
• Added README.md and CHANGELOG.md per spec-ocas-skill-publishing.md.

v1.0.1 — Prior

• Pre-compliance release.