beatra-ai/beatra-skills/ai-image-realism

Claude Code and agent skill by beatra-ai/beatra-skills

Trust tier 2
  • community
LLM review pending
Source check pending

Checking that the SKILL.md file is still in the repository.

Improve AI image realism without starting over. This focused AI image editor and humanizer targets visible AI image artifacts—plastic skin, malformed hands or faces, extra fingers, repeated textures, and inconsistent lighting—with focused AI image retouching through a local repair or natural whole-image refinement. Use it to make AI portraits, product images, marketing graphics, and social covers look more realistic, remove the obvious AI look, fix AI hands and extra fingers, or refine product imagery, while aiming to keep identity, product shape, brand details, and composition recognizable where possible.

Install

npx vskill@latest install beatra-ai/beatra-skills/ai-image-realism

SKILL.md

The same SKILL.md was published earlier as beatra-ai/beatra-skills/ai-image-realism.

AI Image Realism & Artifact Fixer

Repair the most visible artificial-looking defect in an existing image while keeping the parts the user already accepts recognizable where possible. The host Agent must be able to inspect the source image; uploading a file is transport, not image diagnosis, and this Skill has no separate Beatra image-analysis call.

Best-fit scenes and preservation priorities

Use this workflow to improve an existing portrait, product image, marketing visual, or social cover with plastic skin, malformed hands or faces, repeated textures, inconsistent light, or another visible generated-image artifact. Start from the user's selected image and repair the highest-impact issue without unnecessarily changing the rest.

Treat text, logo geometry, identity, product shape, packaging, and composition named by the user as must-keep priorities. Prefer the smallest local repair that can address the issue, then compare the result with the source and report any visible drift. Use a reinterpretation route only when the user asks for a new scene or composition; never silently replace an edit request with a newly generated image.

Inputs and defaults

The only hard input is one existing image the host Agent can inspect. Reuse the user's destination, named defect, must-keep details, and accepted composition when already known. Ask only for missing information that changes the repair route or paid payload.

Unless the user asks otherwise:

  • fix one highest-impact problem cluster per pass;
  • use model=auto and count=1;
  • preserve the source aspect ratio for edits;
  • avoid changing unrelated areas;
  • use one stable client_request_id for one logical paid request.

Golden path

  1. Inspect the visible source. Compare anatomy, skin and materials, light, background repetition, composition, text, branding, and destination fit. If the host cannot inspect the image, request an accessible file and stop.
  2. Diagnose one priority problem. Use diagnosis and routing when several defects compete or preservation constraints conflict.
  3. Choose the smallest useful repair.
  • Use beatra.images.edit with one or two normalized regions for bounded local defects.
  • Use beatra.images.edit without edit_regions for image-wide texture, skin, sharpness, or light.
  • Use beatra.images.transform for an explicitly requested new scene or composition. Reuse the user's stated must-keeps, set the intended canvas explicitly, and ask only when missing information would materially change the result or paid payload; never inherit its 16:9 default by accident.
  1. Preview and confirm. Show the problem to fix, route, canvas, must-keeps, output count, and one paid-execution boundary. Follow repair recipes for the exact payload. Query beatra.models.list only when a live capability, availability, count, or price decision matters.
  2. Upload and execute once. Use the bundled scripts/mcp_client.py to upload a local file and call the selected MCP tool. Do not configure or call a host Beatra Connector and do not use REST/OpenAPI as a fallback. Create the request identifier only after the payload is final, then submit once.
  3. Poll and review. Follow the same task with beatra.tasks.get; slow polling is never a reason to resubmit. Inspect the returned result when the host can view it. If it cannot, say visual review is incomplete instead of inventing a conclusion.
  4. Deliver or recover. Present every returned artifact or link, actual dimensions, final billing.net_charged_credits, task ID, whether the priority defect improved, and any must-keep drift. Use review and recovery for one focused unexecuted revision. A changed payload requires a new identifier and fresh confirmation.

Decisions that require confirmation

Confirm before a paid call, before changing the composition or canvas, before accepting weaker preservation, before increasing output count, or before replacing model=auto with a chosen paid model. If a task remains non-terminal after the bounded polling window, return its task ID and current state for later recovery rather than creating another task.

If the tool returns insufficient credits: give the user the tool message in full. Do not change the URL. You may translate the other words. Do not replace it with your own top-up copy. Point the user to pay using that message.

When the user asks how many credits remain, call beatra.wallet.get. When they ask what was charged, call beatra.wallet.ledger. Both are read-only. Do not invent an account-balance or top-up tool. Do not make wallet.get a required step before every paid submit.

References by task

Updates and security

This Skill runs only the files installed with it and leaves them unchanged while it runs; it never downloads or executes other code. It talks to Beatra's official services over HTTPS, and a file you choose to upload goes only to the one-time upload address Beatra issues for it.

Once a day at most, the bundled client reads Beatra's public version list. When the store or tool you installed from has a newer version, it prints one line with the command that updates it there. Updating stays your choice, through that store or tool. See updates and security.

Security

  • Security scanPassed, 100 of 100Pattern checks on the skill's files.
  • Model reviewNot runA model review is scheduled for the next rescan.

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

Repository stars
2
7-day trend score
10

Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/beatra-ai/beatra-skills/ai-image-realism/badge)](https://verified-skill.com/skills/beatra-ai/beatra-skills/ai-image-realism)

Something wrong with this skill or its page? Tell us.