Ohmycaptcha
Apps & AutomationohmycaptchaOhmycaptcha is an Agent Skill for apps & automation published in the shenhao-stu/ohmycaptcha repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.
Deploy, configure, validate, and integrate the OhMyCaptcha captcha-solving service. Use when working with YesCaptcha-style APIs, flow2api integration, reCAPTCHA/hCaptcha/Turnstile task creation, image classification, SGLang local model deployment, Render/Hugging Face cloud deployment, or OpenAI-compatible multimodal model setup. Also use when the user asks how to self-host a captcha-solving service or wants request/response examples for OhMyCaptcha.
How to use this skill
Bring this guide into your coding agent with a prompt tailored to the tool you use.
- Open your project in Codex.
- Copy and paste the prompt below.
- Review files and risks before approving any action.
Inspect the Agent Skill at https://github.com/shenhao-stu/ohmycaptcha/blob/HEAD/skills/ohmycaptcha/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/ohmycaptcha, but wait for my approval before writing files or running scripts.
Copying this prompt does not install or run the skill. Review third-party files before use.
OhMyCaptcha Skill
Operational guidance for deploying and integrating the OhMyCaptcha service.
Model architecture
OhMyCaptcha uses two model backends:
- Local model — self-hosted via SGLang/vLLM (e.g.
Qwen/Qwen3.5-2B). Handles image recognition and classification tasks. Configured viaLOCAL_BASE_URL,LOCAL_API_KEY,LOCAL_MODEL. - Cloud model — remote OpenAI-compatible API (e.g.
gpt-5.4). Handles audio transcription and complex reasoning. Configured viaCLOUD_BASE_URL,CLOUD_API_KEY,CLOUD_MODEL.
Supported task types (19 total)
Browser-based (12)
RecaptchaV3TaskProxyless, RecaptchaV3TaskProxylessM1, RecaptchaV3TaskProxylessM1S7, RecaptchaV3TaskProxylessM1S9, RecaptchaV3EnterpriseTask, RecaptchaV3EnterpriseTaskM1, NoCaptchaTaskProxyless, RecaptchaV2TaskProxyless, RecaptchaV2EnterpriseTaskProxyless, HCaptchaTaskProxyless, TurnstileTaskProxyless, TurnstileTaskProxylessM1
Image recognition (3)
ImageToTextTask, ImageToTextTaskMuggle, ImageToTextTaskM1
Image classification (4)
HCaptchaClassification, ReCaptchaV2Classification, FunCaptchaClassification, AwsClassification
Local model setup (SGLang)
pip install "sglang[all]>=0.4.6.post1"
# From ModelScope (China):
export SGLANG_USE_MODELSCOPE=true
python -m sglang.launch_server --model-path Qwen/Qwen3.5-2B --port 30000
Then configure OhMyCaptcha:
export LOCAL_BASE_URL="http://localhost:30000/v1"
export LOCAL_MODEL="Qwen/Qwen3.5-2B"
Startup checklist
- Install dependencies:
pip install -r requirements.txt && playwright install --with-deps chromium - Start local model server (SGLang on port 30000)
- Set env vars:
LOCAL_BASE_URL,CLOUD_BASE_URL,CLOUD_API_KEY,CLIENT_KEY - Start service:
python main.py - Verify:
curl http://localhost:8000/api/v1/health - Test: create a reCAPTCHA v3 task against
https://antcpt.com/score_detector/with key6LcR_okUAAAAAPYrPe-HK_0RULO1aZM15ENyM-Mf
Response rules
- Prefer the repository's documented behavior over assumptions.
- Use placeholder credentials only. Never expose real secrets.
- Be explicit about limitations:
minScoreis compatibility-only- Task storage is in-memory with 10-min TTL
- reCAPTCHA v2 and hCaptcha may require image classification fallback in headless environments
- For deployment help, reference
docs/deployment/local-model.md,docs/deployment/render.md,docs/deployment/huggingface.md. - For API usage, reference
docs/api-reference.mdand the usage guides underdocs/usage/.