Menu
Use when integrating LLM calls via LiteLLM/Bedrock for evaluation or content generation
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Basé sur la classification professionnelle SOC
Record browser-based feature demos as GIFs via Playwright and optionally document them in docs/features/index.md. Trigger with "Demo <feature>" or "Demo and Document <feature>".
Use when creating service modules in apps/api/skillhub/services/
Use when writing API tests in apps/api/tests/
Use when implementing division-based access control on any endpoint
Use when adding services to docker-compose.yml or creating Dockerfiles
Use when implementing error handling across the API, frontend, or MCP server
| name | llm-integration-builder |
| description | Use when integrating LLM calls via LiteLLM/Bedrock for evaluation or content generation |
import httpx
async def call_llm(prompt: str, content: str) -> dict:
if not settings.LLM_ROUTER_URL:
return {"pass": True, "score": 85, "findings": [], "skipped": True}
response = await httpx.AsyncClient().post(
f"{settings.LLM_ROUTER_URL}/v1/chat/completions",
json={
"model": "bedrock/anthropic.claude-3-5-sonnet-20241022-v2:0",
"messages": [
{"role": "system", "content": prompt},
{"role": "user", "content": content}
]
},
timeout=30.0
)
return parse_verdict(response.json())
flag = db.query(FeatureFlag).filter_by(key="llm_judge_enabled").first()
if not flag or not flag.enabled:
return JudgeVerdict(passed=True, score=85, findings=[], skipped=True)
{"pass": true, "score": 82, "findings": [
{"severity": "medium", "category": "quality", "message": "..."}
], "summary": "..."}
Score >= 70 = pass. Any CRITICAL finding = auto-fail.
apps/api/skillhub/services/llm_judge.pyapps/api/skillhub/services/submission.pyllm_judge_enabled in seed data