compose-creative
starsforks
How to install compose-creative?
indranilbanerjee/socialforge/compose-creative
$npx -y skills add indranilbanerjee/socialforge --skill compose-creativeCore creative engine. Generates images and video using 4 modes with brand assets, AI compositing, and platform resizing.
SKILL.md
| 1 | # /socialforge:compose-creative — Creative Production Engine |
| 2 | |
| 3 | The core production skill. Takes asset matching results and produces visual and video assets for each post according to its assigned creative mode, using a staged human-in-the-loop approval flow. |
| 4 | |
| 5 | ## Context efficiency |
| 6 | |
| 7 | Asset-heavy skill. **Grep before Read** the asset catalog (`${CLAUDE_PLUGIN_DATA}/<brand>/assets/index.json`) — never list the asset directory. Reference generated images / videos by path, not by loading metadata. Brand profile loads once per session. |
| 8 | |
| 9 | ## The 4 Creative Modes |
| 10 | |
| 11 | ### MODE 1: ANCHOR_COMPOSE |
| 12 | Brand asset is the untouchable center. AI generates scene around it. |
| 13 | 1. Remove background from brand asset (rembg) |
| 14 | 2. Generate surrounding scene via AI using brand style references |
| 15 | 3. Composite asset onto generated scene (Pillow) |
| 16 | 4. Add shadow/reflection for natural placement |
| 17 | 5. Add text overlay + logo |
| 18 | 6. Resize per platform |
| 19 | |
| 20 | ### MODE 2: ENHANCE_EXTEND |
| 21 | Brand asset is the foundation. AI modifies periphery only. |
| 22 | 1. Feed real image to AI edit endpoint |
| 23 | 2. AI extends/enhances while preserving core subject |
| 24 | 3. Verify core subject preservation |
| 25 | 4. Add overlay + resize |
| 26 | |
| 27 | ### MODE 3: STYLE_REFERENCED |
| 28 | No brand asset composited. Style reference photos guide AI generation. |
| 29 | 1. Load 2-8 brand style reference images |
| 30 | 2. Construct 5-layer prompt (brand identity, post context, creative direction, image rules, technical) |
| 31 | 3. Feed references + prompt to AI generation |
| 32 | 4. Quality review — does output match brand DNA? |
| 33 | 5. Add overlay + resize |
| 34 | |
| 35 | ### MODE 4: PURE_CREATIVE |
| 36 | Full AI generation with only text prompt + brand colors/mood. |
| 37 | 1. Construct prompt from brief + brand-config visual style |
| 38 | 2. Generate (text-only, no reference images) |
| 39 | 3. Quality review |
| 40 | 4. Add overlay + resize |
| 41 | |
| 42 | ## Prerequisites |
| 43 | |
| 44 | - `asset-matches.json` must exist (generated by `/socialforge:match-assets` -> match_assets.py) |
| 45 | - `calendar-data.json` must exist (generated by `/socialforge:parse-calendar`) |
| 46 | - `brand-config.json` must exist for active brand |
| 47 | - `cost-log.json` initialized (via status_manager.py --action init-month) |
| 48 | - Valid credentials for Gemini via Vertex AI (images) and WaveSpeed / Kling v3.0 (video) via credential_manager.py |
| 49 | |
| 50 | If any prerequisite is missing: "Run `/socialforge:match-assets` first — creative production needs asset matching results." |
| 51 | |
| 52 | ## Process (For Each IMAGE Post) — 4-Stage Approval Flow |
| 53 | |
| 54 | ### Stage 1: Creative Direction (No API Call) |
| 55 | Present 2-3 creative direction options to the user. Each option includes: |
| 56 | - A description of the visual concept |
| 57 | - Mood, color palette, and composition approach |
| 58 | - How the brand asset will be used (per the assigned creative mode) |
| 59 | - Rough reference to style/tone |
| 60 | |
| 61 | No image generation happens here. The user picks a direction before any credits are spent. |
| 62 | |
| 63 | ### Stage 2: Confirm Details |
| 64 | Once the user selects a direction, confirm the specifics before generating: |
| 65 | - Final prompt details and any refinements the user wants |
| 66 | - Aspect ratio and platform targets |
| 67 | - Text overlay copy and placement preferences |
| 68 | - Any special instructions (e.g., "warmer tones", "more negative space") |
| 69 | |
| 70 | User explicitly approves to proceed to generation. |
| 71 | |
| 72 | ### Stage 3: Generate and Select (Gemini via Vertex AI) |
| 73 | - Generate 2-3 image versions based on the confirmed direction |
| 74 | - Show all versions inline via the Read tool so the user can compare |
| 75 | - User picks their preferred version (or requests another round with adjustments) |
| 76 | - Max 3 generation rounds per post before escalating |
| 77 | |
| 78 | ### Stage 4: Post-Process and Save |
| 79 | After the user selects their preferred image: |
| 80 | 1. Apply text overlay via compose_text_overlay.py |
| 81 | 2. Apply logo per brand-config.json (position, opacity, size, platform exclusions) |
| 82 | 3. Resize per platform via resize_image.py |
| 83 | 4. Save final assets to the post-specific folder: `{post_folder}/final/` |
| 84 | 5. Update status-tracker.json |
| 85 | 6. Log prompt to `shared/prompt-logs/` |
| 86 | 7. Track API cost in cost-log.json |
| 87 | |
| 88 | ## Process (For Each VIDEO Post) — 5-Stage Approval Flow |
| 89 | |
| 90 | ### Stage 1: Video Concept (No API Call) |
| 91 | Present 2-3 video concept ideas to the user. Each option includes: |
| 92 | - A narrative arc or visual concept for the short-form video |
| 93 | - Description of the opening frame, motion/transition, and closing frame |
| 94 | - Mood, pacing, and style notes |
| 95 | - How brand assets or style references will be incorporated |
| 96 | |
| 97 | User picks a concept before any generation begins. |
| 98 | |
| 99 | ### Stage 2: First Frame Generation (Gemini via Vertex AI) |
| 100 | - Generate 2 first-frame options based on the selected concept |
| 101 | - Show both inline via the Read tool |
| 102 | - User picks their preferred first frame (or requests adjustments) |
| 103 | |
| 104 | ### Stage 3: Last Frame Generation (Gemini via Vertex AI) |
| 105 | - Generate 2 last-frame options that complement the approved first frame |
| 106 | - Show both inline via the Read tool |
| 107 | - User pick |
People who liked this also viewed
6 picksType
- seo-auditWhen the user wants to audit, review, or diagnose SEO issues on their site. Also use when the user mentions "SEO audit," "technical SEO," "why am I not ranking," "SEO issues," "on-page SEO," "meta tagSkillsJul 2026
- copywritingWhen the user wants to write, rewrite, or improve marketing copy for any page — including homepage, landing pages, pricing pages, feature pages, about pages, or product pages. Also use when the user sSkillsJul 2026
- marketing-psychologyWhen the user wants to apply psychological principles, mental models, or behavioral science to marketing. Also use when the user mentions 'psychology,' 'mental models,' 'cognitive bias,' 'persuasion,'SkillsJul 2026
- content-strategyWhen the user wants to plan a content strategy, decide what content to create, or figure out what topics to cover. Also use when the user mentions "content strategy," "what should I write about," "conSkillsJul 2026
- programmatic-seoWhen the user wants to create SEO-driven pages at scale using templates and data. Also use when the user mentions "programmatic SEO," "template pages," "pages at scale," "directory pages," "location pSkillsJul 2026
- marketing-ideasWhen the user needs marketing ideas, inspiration, or strategies for their SaaS or software product. Also use when the user asks for 'marketing ideas,' 'growth ideas,' 'how to market,' 'marketing stratSkillsJul 2026
