Start with the prompt you would otherwise write:<\/p>\n\n\n\n
Review this campaign report and tell me what to do.\n<\/code><\/pre>\n\n\n\nIt leaves the evidence boundary, calculation method, output, and stopping rule undefined. A more useful version is:<\/p>\n\n\n\n
Use only the supplied campaign rows. Calculate CPC, CPL, and qualified CPL as\nspend divided by the named denominator. Return one table with numerators and\ndenominators, then two sentences limited to differences visible in these rows.\nIf a field is missing, write UNKNOWN. Do not infer revenue, incrementality, or\nbudget recommendations. Stop after the table and conclusion.\n<\/code><\/pre>\n\n\n\nThe second prompt is not longer for its own sake. It turns \u201ctell me what to do\u201d into a result a reviewer can check. This guide shows how to make that kind of improvement repeatable, reduce wasted context and retries, and evaluate changes without promising that shorter prompts are always cheaper or equally accurate. The examples use supplied miniature inputs and manually written ideal outputs; they are illustrative contracts, not live Astra runs, benchmark results, or customer case studies. For broader model context, see DMT\u2019s Astra overview<\/a>; this guide stays focused on prompting.<\/p>\n\n\n\nThe prompt contract: seven decisions<\/h2>\n\n\n\n
Astra can reason across a large source set, use tools, and produce structured output. Those capabilities do not remove the need to define the job. A practical contract names the outcome, evidence, method, prohibitions, answer shape, uncertainty rule, and autonomy\/stop boundary; the compact template below combines those decisions.<\/p>\n\n\n\n
OpenAI\u2019s latest-model guidance<\/a> describes Astra behavior that matters for prompt design: it may ask for clarification when missing input could change the result, it tends toward detailed formatted responses, it can be sensitive to instructions in accessible skills or files, and coding tasks can trigger broader testing than a small change needs. Prompt for the initiative, detail, delegation, and testing scope you actually want; do not assume a generic \u201cbe helpful\u201d instruction will choose correctly for you.<\/p>\n\n\n\nA compact template<\/h3>\n\n\n\n
Use this contract for a new task. The labels are not magic syntax; they make omissions visible to a human reviewer.<\/p>\n\n\n\n
ROLE\nYou are the [role] supporting [decision or workflow].\n\nOUTCOME\nProduce [specific deliverable] for [audience]. Success means [observable acceptance condition].\n\nEVIDENCE\nUse only the supplied source pack unless I explicitly permit outside research.\nFor every material claim, include a source ID and quote or locator. If the pack is silent, write UNKNOWN.\n\nMETHOD\n1. Extract facts and conflicts.\n2. Separate fact, inference, recommendation, and open question.\n3. Produce the requested deliverable.\n\nCONSTRAINTS\nKeep [search question \/ schema \/ tone \/ length \/ date boundary]. Do not invent [metrics, availability, sources, or results].\n\nOUTPUT\nReturn [exact headings or JSON fields]. Use [units]. Keep [field] to [limit].\n\nAUTONOMY AND STOP\nYou may [read-only actions]. Do not [writes or external side effects].\nStop and ask one focused question when [specific ambiguity]. Otherwise make the smallest stated assumption and label it.\n<\/code><\/pre>\n\n\n\nThe contract is deliberately explicit about evidence and stopping. A long prompt with no acceptance condition is still vague. A short prompt with a clear input, output, and failure rule can be easier to evaluate.<\/p>\n\n\n\n
Worked example 1: a source-backed research brief<\/h2>\n\n\n\n
The following miniature packet is fictional and deliberately small. It shows how to make a missing fact visible instead of letting a fluent answer fill the gap.<\/p>\n\n\n\n
Supplied input<\/h3>\n\n\n\nSOURCE S1 \u2014 release note, 2026-09-01\n\u201cFeature Atlas is available to Pro accounts in the web application.\u201d\n\nSOURCE S2 \u2014 workspace note, 2026-09-02\n\u201cBusiness workspaces require an administrator to enable Atlas before members can select it.\u201d\n\nQUESTION\nWrite a 60\u201390-word availability note for a small-business reader. Do not infer region, API, or mobile availability.\n<\/code><\/pre>\n\n\n\nPrompt<\/h3>\n\n\n\nYou are an evidence editor. Use only S1 and S2.\nReturn JSON with: answer, supported_claims, unknowns, and action.\nEach supported_claim has source_id and claim. Put any detail not stated in S1 or S2 in unknowns.\nThe answer must be 60\u201390 words, direct, and must not imply that web availability proves API or mobile availability.\nIf the two sources conflict, return the conflict instead of choosing a side.\n<\/code><\/pre>\n\n\n\nIdeal output for the supplied packet<\/h3>\n\n\n\n{\n "answer": "Feature Atlas is available to Pro accounts in the web application. In a Business workspace, an administrator must enable Atlas before members can select it. The supplied notes do not establish API access, mobile availability, regional coverage, rollout timing, or eligibility for other plans. Obtain a current official source covering all those unknown details before promising access to a customer.",\n "supported_claims": [\n {"source_id": "S1", "claim": "Atlas is available to Pro accounts in the web application."},\n {"source_id": "S2", "claim": "Business workspaces require administrator enablement."}\n ],\n "unknowns": ["API access", "mobile availability", "regions", "other plans", "rollout timing"],\n "action": "Obtain a current official source for the unknowns before promising access."\n}\n<\/code><\/pre>\n\n\n\nThe useful behavior is not the prose style. It is the boundary around the evidence. If a later source adds API access, the prompt can accept it only when the source pack is updated. A source ID also makes an editorial review faster than asking a reviewer to rediscover which sentence came from which file.<\/p>\n\n\n\n
For a real research workflow, preserve a source manifest outside the model call with URL, retrieval date, content hash, and access restrictions. Ask Astra to cite the manifest\u2019s IDs, but let the application validate that each returned ID exists. A model-generated URL is not proof that a page was consulted.<\/p>\n\n\n\n
Worked example 2: tabular analysis with inspectable arithmetic<\/h2>\n\n\n\n
Prompting for a table is often more useful than asking for an attractive narrative. Give Astra the rows, define the formulas, and require it to show enough intermediate values for a reviewer to reproduce the result. The following miniature input is invented for this example.<\/p>\n\n\n\n
Supplied input<\/h3>\n\n\n\n