Claude prompts: 10 copy-ready engineering templates

Can you look at the code below and tell me what it does,
then tell me whether it has performance issues,
and also the function above that I asked you about earlier,
you said changing it to async would be enough...

<context>
We are building a high-concurrency scraping service with about 500 requests per second.
</context>

<code>
def fetch_all(urls):
    results = []
    for url in urls:
        results.append(requests.get(url).text)
    return results
</code>

<task>
1. Explain what this code does in one sentence
2. Identify the most serious performance issue
3. Provide a rewrite that fits the concurrency level in the context
</task>

<output_format>
Answer using the task's 1/2/3 numbering. Keep each point under 5 lines.
</output_format>

import anthropic

client = anthropic.Anthropic(
    api_key="sk-your-ClaudeAPI-key",
    base_url="https://gw.claudeapi.com"
)

response = client.messages.create(
    model="claude-sonnet-4-6",
    max_tokens=1024,
    messages=[{"role": "user", "content": prompt_with_xml}]
)

<task>
From a product review, extract the following fields and return JSON:
- sentiment: positive / negative / neutral
- mentioned_features: array
- has_complaint: bool
</task>

<examples>
<example>
<input>The battery life is good, but charging is too slow; it takes 3 hours.</input>
<output>{"sentiment": "neutral", "mentioned_features": ["battery life", "charging speed"], "has_complaint": true}</output>
</example>

<example>
<input>The display looks excellent, with vivid colors.</input>
<output>{"sentiment": "positive", "mentioned_features": ["display"], "has_complaint": false}</output>
</example>

<example>
<input>Support never replied to my message. Bad experience.</input>
<output>{"sentiment": "negative", "mentioned_features": ["support"], "has_complaint": true}</output>
</example>
</examples>

<user_input>
The camera quality is impressive, but startup is a bit slow. Overall, I am satisfied.
</user_input>

<question>
A car travels from A to B in 4 hours at an average speed of 60 km/h. On the return trip along the same route, rain lowers the average speed to 40 km/h. What is the average speed for the round trip?
</question>

<instructions>
Answer using the following structure:

<thinking>
- Step 1: List the known conditions
- Step 2: Calculate the one-way distance
- Step 3: Calculate the return-trip time
- Step 4: Use total distance / total time to calculate the average speed
</thinking>

<answer>
Final answer in km/h, rounded to 1 decimal place
</answer>
</instructions>

response = client.messages.create(
    model="claude-opus-4-7",
    max_tokens=16000,
    thinking={"type": "enabled", "budget_tokens": 8000},
    messages=[{"role": "user", "content": question}]
)

response = client.messages.create(
    model="claude-sonnet-4-6",
    max_tokens=1024,
    messages=[
        {"role": "user", "content": "Return JSON with 3 interesting Python libraries, each containing name and desc."},
        {"role": "assistant", "content": "{\n  \"libraries\": ["}  # Make the model continue from here
    ]
)
print("{\n  \"libraries\": [" + response.content[0].text)

{"role": "assistant", "content": "<legal_advisor_response>\n"}

SYSTEM = """You are a senior Python backend engineer specializing in FastAPI and high-concurrency optimization.

Response rules:
1. Provide directly runnable code only, not pseudocode
2. Avoid multi-line docstrings and filler comments
3. For version-specific details, assume FastAPI 0.115+ and Python 3.12+
4. If the user's request is clearly wrong, point it out directly and provide the correct approach
5. If you are unsure, say "I am not sure" instead of inventing an answer
"""

response = client.messages.create(
    model="claude-sonnet-4-6",
    max_tokens=2048,
    system=SYSTEM,
    messages=[{"role": "user", "content": user_query}]
)

Answer in no more than 100 words.

Answer using the following structure. Each item must be 1 sentence:
- Conclusion:
- Main reason:
- Recommended approach:

messages=[
    {"role": "user", "content": "Summarize the article above in one sentence."},
    {"role": "assistant", "content": "One-sentence summary:"}
],
max_tokens=80

When answering the following question, if any of the conditions below apply, say "I cannot answer this accurately" and stop:

- Specific data required for the calculation is missing
- The answer depends on a specific legal provision and you are not sure about the latest version
- The request involves a specific medical diagnosis
- The answer depends on facts you cannot verify

prompt = f"""<document>
{long_document}
</document>

Answer based on the document above: {question}

If the document does not contain relevant information, say "The document does not mention this" and do not guess.
"""

response = client.messages.create(
    model="claude-opus-4-7",
    max_tokens=2048,
    messages=[{"role": "user", "content": prompt}]
)

messages=[
    {"role": "user", "content": [
        {"type": "text", "text": f"<document>\n{long_document}\n</document>",
         "cache_control": {"type": "ephemeral"}},
        {"type": "text", "text": f"Question: {question}"}
    ]}
]

<task>
Write a PR description from the git diff below, matching the team's style.
</task>

<style_examples>
<example>
## What
Add retry logic to the payment webhook handler.

## Why
~3% of webhook calls were failing silently due to upstream timeouts.

## How
Exponential backoff, max 5 retries, with DLQ for permanent failures.

## Test
- Unit: tests/payments/test_webhook_retry.py
- Manual: triggered 50 mock failures locally
</example>

<example>
## What
Switch user search from LIKE to Postgres tsvector.

## Why
P99 latency on /api/users/search was 1.8s; reduces to <100ms.

## How
Added GIN index on `users.search_vector`; migration in 0042.

## Test
- Unit: existing search tests
- Benchmark: scripts/bench/users_search.py before/after
</example>
</style_examples>

<diff>
{git_diff_content}
</diff>

SYSTEM = """You are <role>, <constraints>.

Response rules:
1. <rule 1>
2. <rule 2>
3. <rule 3>
"""

USER_PROMPT = f"""
<context>
{background_information}
</context>

<examples>
<example>
<input>...</input>
<output>...</output>
</example>
</examples>

<task>
{specific_task}
</task>

<output_format>
{format_requirements}
</output_format>
"""

response = client.messages.create(
    model="claude-sonnet-4-6",
    max_tokens=2048,
    system=SYSTEM,
    messages=[
        {"role": "user", "content": USER_PROMPT},
        {"role": "assistant", "content": "<format-locking prefill>"}
    ]
)