image-prompt-portinglisted

Port an external photoreal image prompt (a viral JSON prompt from a tweet, another creator's "photorealistic" template, a prompt pasted by the user) into your own image pipeline, or add a new shot type (camera angle, framing, face shot) to it. Keeps the structure that makes these prompts work (subject, pose, wardrobe, setting, light, camera, realism, micro-details, negatives, priorities, output), replaces the domain content, adds your invariants (no readable text or logos, camera never visible, one light source, calm zone for overlay text), keeps the JSON valid after placeholder filling, keeps prompts in files instead of code, and walks the checklist so the new shot type is actually used by the pipeline. Use it whenever the user shares an image prompt or a link to one and says "take inspiration", "rework ours like this", "use this style", or asks for a new angle, framing or photo style for carousels or ads, even if they never say "prompt" or "shot type".
josephbedminster/indie-marketing-skills · ★ 1 · AI & Automation · score 74

Install: claude install-skill josephbedminster/indie-marketing-skills

# Porting an image prompt into your pipeline What makes a good shared photoreal prompt work is its **structure**, not its content. A JSON prompt that separates subject, pose, wardrobe, setting, light, camera, realism requirements, fine details, negatives, a priority order and the output spec is followed far more literally by GPT-style image models than the same ideas in prose. So you port the skeleton and throw away the gym mirror. ## Before you start Read `.claude/ugc-config.md` (paths of the pipeline, where prompts live, the placeholder filler, the check commands, the shot-type files, brand invariants). Missing? Create it with the questions and template described in the `ai-persona` skill of this plugin (`../ai-persona/references/config-template.md`). Read the **Project conventions** of `.claude/posthog-funnel-map.md` if present. ## Port a prompt 1. **Read the source in full.** If it is a tweet or a page, open it in the browser tool and extract the text; don't work from a screenshot excerpt. 2. **Keep the structure, replace all domain content.** Typical top-level blocks worth keeping: `type`, `objective`, `image_format`, `subject` (count, identity, attractiveness, skin, hair, face, hands, clothing), `shot` (kind, camera position, pose, gaze, expression, subject position, lens), `environment`, `text_overlay_composition`, `lighting`, `camera_characteristics`, `color_palette`, `realism_requirements`, `fine_details`, `negative_prompt`, `identity_handli