Project CLAUDE.md & aik custom Cerebras skill

Ab human element ka daur. Aik line ka prompt kuch bana dega, lekin agar humein aik mazboot, differentiated product chahiye, to yeh woh jagah hai jahan hum apni fingerprints lagate hain — aik ache se likhe gaye project CLAUDE.md aur apne khud ke skill ke zariye. Dono likhte hain.

01Project CLAUDE.md: scene set karein

Repo root mein aik naya CLAUDE.md kholein aur product ko high level par describe karein — detailed requirements nahi (woh Jira mein hain), bas itna ke Claude coherent rahe kyunke use game pata hai:

Context ki aik line add karein — “the initial implementation is a front-end-only prototype” — taake Claude starting state samajh sake.

02Development-process section

Bata dein ke kaam kaise flow hona chahiye, taake har ticket usi disciplined tareeqe se banaye:

Use your Atlassian tools to read the feature instructions from Jira.
Develop the feature — do not skip any steps.
Thoroughly test with unit tests and integration tests; fix any issues.
Submit a PR using your GitHub tools.

03AI-design section (Cerebras enter karta hai)

Kyunke yeh product LLM ko call karta hai user ko samajhne aur fields fill karne ke liye, hum batatay hain kaise:

When writing code to call an LLM, use your cerebras skill:
LiteLLM via OpenRouter to the `openai/gpt-oss-120b` model,
with Cerebras as the inference provider.
Use structured output so results can populate the document fields.

“Use your cerebras skill” — aik skill jo hum abhi banana walay hain. Yeh skills ka fayda hai: CLAUDE.md chhota rehta hai aur sirf expertise ki taraf point karta hai.

04Technical-design section

Architecture guardrails dein taake Claude kuch coherent aur containerised banaye:

Package the whole project into a Docker container.
Backend in `backend/` — a UV project using FastAPI.
Frontend in `frontend/`.
Provide scripts to start and stop the app.
Use SQLite, created fresh each time the container starts,
with a users table supporting sign-up and sign-in.

05Cerebras skill banayein

Cerebras kyun? Humein chahiye ke AI bohot tezi se jawab de. Hum aik open model use karte hain, gpt-oss-120b, OpenRouter par, lekin provider ko Cerebras par pin karte hain — aap cheapest option se thora zyada pay karte hain, lekin responses almost instantly wapas aate hain. (Free models pasand hain? Bas yeh skip kar dein ya skill ko kahin aur point kar dein.)

Folder .claude/skills/cerebras/ banayein aur andar SKILL.md rakhen. Bilkul top par metadata block ko yeh exact format use karna chahiye — jaise API key, galat kiya to kaam nahi karega:

---
name: cerebras
description: Cerebras inference. Use this to write code to call an
  LLM using LiteLLM and OpenRouter with the Cerebras inference provider.
---

# Calling an LLM via Cerebras
These instructions let you write code to call an LLM with Cerebras
specified as the inference provider. Read the model + key from `.env`.

# structured output, pinned to the Cerebras provider
resp = litellm.completion(
    model="openrouter/openai/gpt-oss-120b",
    messages=messages,
    api_key=os.environ["OPENROUTER_API_KEY"],
    response_format=NDAFields,          # a Pydantic schema
    extra_body={"provider": {"order": ["Cerebras"],
                          "allow_fallbacks": False}},
)

06Key wire karein aur chalo

Skill ko OpenRouter API key ki zaroorat hai. Apni existing .env copy kar lein — aur .gitignore ki wajah se jo humne set kiya, yeh kabhi commit nahi hogi:

cp ../pm/.env .   # brings your OPENROUTER_API_KEY into this project

Akhir mein, CLAUDE.md mein bilkul clear kar dein: “There is an OpenRouter API key in the .env file in the project root.” Ab hamare paas aik project CLAUDE.md, aik home CLAUDE.md, aik .env, aur aik brand-new Cerebras skill hai — product poori tarah se build ke liye primed hai.