CoreWise Academy

Prompting & Context · Layer II / Broad

Brief the model like a brilliant new hire

Claude is a brilliant new hire with no context on your job. Brief it like one: explicit instructions with motivation, examples, XML tags, roles, and context ordering.

Nº 014 · Vol. I·6 min read· Updated July 2026

“a colleague who is confused, Claude will be too”

The fastest way to get a bad answer is to prompt as if the model already knows what you know. It does not. Anthropic’s prompting guide frames every current model (Fable 5, Opus 4.8, Sonnet 5, and the rest) as a brilliant new employee: sharp, fast, capable of excellent work, and holding zero context on your job, your conventions, or what “good” means here. The golden rule falls straight out of that. If a colleague with minimal context would be confused by your prompt, Claude will be too. This is the first course in prompting, and it teaches the parts that do not change when the model does. (OpenAI’s guidance points the same way; the overlap is mapped in Claude and OpenAI, one prompting playbook.)

01Say the quiet part

Vague prompts fail because the model cannot infer what you never stated. If you want the above-and-beyond version, ask for it. If you have a format in mind, describe it. The colleague test catches most of this before you hit send: read your prompt as someone who just joined and knows nothing about the task, and every place they would stop and ask a question is a place Claude will guess. The test takes 30 seconds and pays for itself on the first prompt.

“If a colleague with minimal context would be confused by your prompt, Claude will be too.”

The sharper move is to give the reason behind a rule, not just the rule. “NEVER use ellipses” is a bare ban that covers one symbol and nothing else. “Your output will be read aloud by a text-to-speech engine, so avoid anything that does not read cleanly when spoken” is a motivation, and the model generalizes from it, dropping the ellipses, yes, but also the emoji, the stray markdown, and the table that would come out as noise. State the why and Claude extends the rule to cases you never listed. That single habit does more work than 10 times as many bans. If your prompts already lean hard on forceful bans and the model over-corrects, Stop shouting at the model shows how to dial that language back.

02Show, tag, and give the model a role

Three techniques carry across every model. First, examples. Wrap three to five in <example> tags inside an <examples> block, and make them relevant and diverse rather than five near-copies. The model learns your desired output far better from demonstrations than from description, and diversity is what stops it latching onto one shape.

Second, tags. When a prompt mixes instructions, context, and the thing to act on, XML tags stop Claude from confusing one for another. Name them for what they hold (<instructions>, <context>, <input>), keep the names consistent, and nest for hierarchy. The tags are cheap and the separation is real: a document that contains its own instructions will not get mistaken for yours if it sits inside <input>.

Third, a role. One sentence in the system parameter, “You are a helpful coding assistant specializing in Python”, measurably focuses tone and behavior. The system parameter is the right home for standing character; the user turn is for the task.

03Order the docs, mean every word

For inputs past 20k tokens, put the longform documents at the top and the query at the very end. Anthropic reports end-placed queries improved response quality by up to 30 percent on complex multidocument tests, which is a large return for 5 minutes of rearranging text you already have. Give each document its own <document> tag with <document_content> and <source> subtags, and for work that leans on pulling facts from sources, ask Claude to pull relevant quotes into <quotes> tags before it answers. That lets the model trace each conclusion back to its source.

The last fundamental is the one people trip over most: the latest models do what you literally say. “Can you suggest some changes” returns suggestions, not edits, because that is what you asked for. Want the edit made? Use the imperative: “Change this function to handle the null case.” The same literalness governs formatting, and it steers best when you say what to do rather than what to avoid. “Write in flowing prose paragraphs” beats “don’t use bullet points,” and if you want less markdown in the output, strip the markdown from your own prompt, because the model matches the style it is handed. Tell it the target, not the sin.

“Write in flowing prose paragraphs beats don’t use bullet points: say what to do, not what to avoid.”

The fundamentals, side by side:

FundamentalThe moveWhy it holds
Colleague testRead the prompt as a context-free coworkerEvery unasked question becomes a guess
MotivationState the why behind each ruleThe model generalizes past the named case
Examples3 to 5 diverse, tagged demonstrationsDemonstrations beat description
OrderingDocuments on top, query at the endUp to 30 percent better on multidocument tests
LiteralnessImperative phrasing, positive steeringThe model does what you literally say

Further reading

FAQ

Why does "your output will be read aloud by a text-to-speech engine" beat "NEVER use ellipses" as an instruction?

The bare ban covers exactly one symbol. The motivation tells Claude the underlying goal, text that sounds right when spoken, so it also drops the emoji, the stray asterisks, and the table that would read as noise. State the why and the model generalizes the rule to cases you never listed.

In a 40k-token prompt with six documents and a question, where does the question go, and why?

At the very end, after all six documents. Anthropic's tests on complex multidocument inputs showed end-placed queries improved response quality by up to 30 percent. Put the longform material at the top, wrap each document with its own source tag, and close with the instruction.

You typed "Can you suggest some changes to this function?" and Claude described edits instead of making them. What went wrong?

Nothing, from the model's side. Latest Claude models follow instructions literally, and you asked for suggestions, so you got suggestions. Use the imperative ("Change this function to...") when you want the edit performed rather than proposed.

For your agent

A distilled, agent-ready version of this guide, not a skill. Copy it below, or point the agent at corewise.academy/guides/brief-the-model.md.

file · guide.md brief-the-model 49 lines · markdown
# Brief the model like a brilliant new hire Consult when writing or editing a prompt and the output is vague, off-format, or misses context you never stated. These are the prompt fundamentals that hold across every current Claude model. The model is a sharp new employee with zero context on your job, your conventions, or what "good" means here. Distilled from the CoreWise guide `brief-the-model`, whose source is Anthropic's Claude prompting best practices. For live model ids, use the `claude-api` skill rather than restating version-pinned names. ## Run the colleague test If a coworker with minimal context would be confused by the prompt, the model will be too. Read the prompt as someone who just joined and knows nothing. Every spot they would stop and ask a question is a spot the model guesses. - Want the above-and-beyond version, ask for it. Have a format in mind, describe it. The model cannot infer what you never stated. ## State the motivation, not just the rule Give the reason behind a rule, not the bare rule. The model generalizes from a motivation to cases you never listed. - "NEVER use ellipses" covers one symbol. "Your output will be read aloud by a text-to-speech engine, so avoid anything that does not read cleanly when spoken" also drops the emoji, stray markdown, and tables that read as noise.- If prompts lean on forceful bans and the model over-corrects, see `stop-shouting-at-the-model` (agent guide: `prompting-newest-claude-models.md`).

+ 33 more lines · the copy includes all of them

then save as .claude/reference/brief-the-model.md
Sources

Original work by the editor, no video source.