Reusable artifact instructions define a recurring reader task, fresh input requirements and review checks without carrying old facts into a new edition.
You review the facts and access. Nothing is published without your approval.
The method behind the reusable artifact instruction sheetAsk your agent to separate fixed task rules from current inputs, then test the instructions against a second edition with a deliberate gap.
How do I create reusable instructions for recurring artifacts?
Describe the recurring reader task, required source inputs, output structure and checks that every edition must pass. Keep changing facts such as dates and numbers in a separate input section. Reference your saved style separately. Test the instructions on a second source bundle, including missing information, before relying on them.
What to give your agent
A recurring audience question and an approved example of a useful result.
The fresh input checklist: reporting period, source dates, definitions and evidence.
Required sections, limits, factual review checks and publication boundaries.
Optional saved style plus where you will store and deliberately reuse the instructions.
Extract the job, not the old content
Describe what the reader needs to learn or decide in each edition. Preserve useful structure and checks, but remove old names, dates, values and conclusions from the instructions. A previous report can be an example of shape without becoming evidence for the next report.
Separate fixed rules from fresh inputs
Use clear sections for the recurring task, current source bundle, optional references and review requirements. Mark which inputs are essential and what to do when they are missing. A prompt with placeholders is not a connection to the source system; the user or authorized agent still provides current data.
Keep style and authority separate
Saved style guides typography, palette and tone. Task instructions define the question, evidence and checks. Neither authorizes publication or access changes. Store the instructions where your workflow can deliberately retrieve them; do not assume every client remembers previous conversations.
Test with a different and incomplete edition
Run the instructions on a new source bundle with changed values, then one with an omitted definition or denominator. Check that the agent uses fresh facts and asks about the gap rather than copying the old answer. Revise the instructions when a recurring error becomes visible.
Help me create reusable instructions for [recurring reader task]. Use my authorized example to identify useful structure, required fresh inputs and review checks—not reusable factual conclusions. Separate fixed task rules, current source data, optional references and saved style. Specify what to do when essential facts are missing or contradictory. Test the instructions with a clearly labeled synthetic second edition if I choose, including a missing input. Do not assume persistent memory, automatic retrieval, scheduling or publication. The artifact output should be static HTML without scripts, forms or live queries. Ask only for missing essentials from existing authorized context. Show the instruction sheet and test results for review. Do not publish or change access without separate explicit approval.
Check before you shareAccuracy, clarity, and permission
Are dates, values and conclusions kept out of the fixed instructions?
Does each new edition require current source material and definitions?
Does a missing-input test produce a question or limitation rather than an invented answer?
Are style reuse, storage, scheduling and publish approval treated as separate concerns?
When HTML fits
A portable instruction sheet for repeat creation, with an input checklist and observable review tests.
When another format fits better
This is not an installed agent skill, automatic memory, scheduler or data connector. Use your client's actual supported storage and scheduling workflow when needed.
What we started withThe same facts, as a wiki page
Fictional team wiki
Cedar monthly service briefing — reusable instructions and test cases
Recurring task
September 7, 2026. This is an invented instruction-testing exercise, not performance data. Explain current-month request completion with closed count, eligible count and rate where both counts are available: closed / eligible × 100.
Fixed rules and style
Use only the current edition’s authorized bundle. State period and definitions, show calculations where possible, ask for missing inputs, and never fill a gap with previous-period values. Do not infer causes or compare incompatible definitions. Use readable headings and one restrained accent; style does not determine arithmetic or permission to share.
Test inputs and expected results
August A1 has 18 closed / 24 eligible, with matching August definitions: 18 / 24 × 100 = 75%. September S1 has 21 closed but no eligible count. A September rate or rate comparison cannot be established; 21 / 24 would misuse August’s denominator. Ask for September’s eligible count, without inventing a deadline or claiming improvement.
Review and setup boundaries
The creator reviews period, definitions, arithmetic, gaps and audience. Show these instructions and both test outcomes in a static reading page. Saving the instructions does not configure memory, scheduling, access or publication; no client installation is requested.
Inspect the HTMLOne self-contained file
The example is sanitized and isolated from this site. No scripts, forms, or live data.
<!doctype html><html lang="en"><head><meta charset="utf-8"><meta name="viewport" content="width=device-width, initial-scale=1"><title>Cedar monthly brief: instructions tested on two editions</title><style>
*{box-sizing:border-box}html{color-scheme:light}body{margin:0;background:#faf6fa;color:#302439;font:17px/1.65 system-ui,sans-serif}main{max-width:1000px;margin:auto;padding:clamp(20px,5vw,64px)}h1,h2,h3,p{margin:0 0 18px}h1{font:500 clamp(32px,5vw,58px)/1.14 Georgia,serif;letter-spacing:-.025em;overflow-wrap:anywhere}h2{font:500 29px/1.3 Georgia,serif;overflow-wrap:anywhere}h3{font-size:19px;line-height:1.4}.label{font:700 12px/1.6 system-ui,sans-serif;letter-spacing:.08em;text-transform:uppercase}.intro{font-size:20px}.lead{background:#53355f;color:white;padding:24px;margin:28px 0}.lead p:last-child{margin:0}.grid{display:grid;grid-template-columns:repeat(auto-fit,minmax(min(100%,230px),1fr));gap:20px;margin:24px 0}.card{background:white;border:1px solid #bbc5c4;padding:24px}.metric{font:500 38px/1.2 Georgia,serif;display:block;margin-bottom:12px}section{margin:36px 0}ul,ol{padding-left:22px}li{margin-bottom:12px}.note{font-size:14px}.warning{border-left:5px solid #956118;padding-left:20px}a{color:inherit;text-underline-offset:4px}a:focus-visible,summary:focus-visible{outline:3px solid #956118;outline-offset:4px}table{width:100%;border-collapse:collapse;table-layout:fixed;font-size:15px}th,td{text-align:left;vertical-align:top;border-bottom:1px solid #bbc5c4;padding:12px 8px;overflow-wrap:anywhere}caption{text-align:left;font-weight:700;padding:12px 0}pre,code{white-space:pre-wrap;overflow-wrap:anywhere}footer{border-top:1px solid #bbc5c4;padding-top:24px;margin-top:40px;font-size:14px}@media(max-width:480px){th,td{padding:10px 4px;font-size:14px}}@media print{body{background:white}.card,.lead{break-inside:avoid}h2,h3{break-after:avoid}}
</style></head><body><main><header><p class="label">Cedar / Fictional instruction test / September 7, 2026</p><h1>Reuse the task. Never reuse a missing denominator.</h1><p class="intro">A portable instruction sheet tested against one complete month and one incomplete month.</p></header><section><h2>The fixed task instructions</h2><div class="card"><ol><li>Use the current month's authorized source bundle only.</li><li>Show the period, closed requests and eligible requests.</li><li>Calculate completion as closed ÷ eligible × 100 only when both counts and their definitions are supplied.</li><li>If an essential input is missing, report what is known and ask for the gap. Do not copy an earlier value or invent a cause.</li><li>Review facts, arithmetic and audience before any separate publication approval.</li></ol><p class="note">Sources T1, F1 and R1. Style is a separate preference, not a calculation rule.</p></div></section><section><h2>Test one: a complete August bundle</h2><div class="lead"><p class="label">A1 · August</p><h3>18 of 24 eligible requests closed</h3><p>18 ÷ 24 × 100 = <strong>75% completion</strong>.</p><p>Both counts cover August and use the same request definition.</p></div></section><section><h2>Test two: an incomplete September bundle</h2><div class="card"><p class="label">S1 · September</p><h3>21 requests closed. Rate unavailable.</h3><p>The eligible request count is missing. Ask for September's eligible count before calculating a completion rate.</p><p class="warning">Do not calculate 21 ÷ 24: 24 belongs to August. These inputs cannot establish that the completion rate improved.</p></div></section><section><h2>What stays fixed—and what must change</h2><h3>Reuse boundary</h3><table><thead><tr><th scope="col">Keep in the instructions</th><th scope="col">Supply for each edition</th></tr></thead><tbody><tr><td>Reader task and rate definition</td><td>Current period and current source counts</td></tr><tr><td>Missing-input behavior and review gate</td><td>Current evidence and unresolved questions</td></tr><tr><td>Readable style preference</td><td>Audience and separate sharing approval</td></tr></tbody></table></section><section><h2>Use deliberately next time</h2><p>Store these instructions in a location supported by your chosen workflow and deliberately provide or retrieve them. This page does not install a skill, remember a future conversation, schedule a report, access a source system or publish an edition.</p><p class="note">Sources V1, R1 and O1. No deadline or causal explanation is supplied.</p></section><footer>Fictional worked example by share/artifacts. Names, dates, figures and records are invented for demonstration. Static snapshot; no live data or automatic updates.</footer></main></body></html>
The original worked scenarioThe source scenario, step by step
A reusable monthly completion brief that catches a missing denominator
Fictional Cedar service updates. Two synthetic editions test the same instructions.
The original notes
Task: explain monthly request completion using closed requests divided by eligible requests. August has 18 closed of 24 eligible. September has 21 closed, eligible count missing. Use current sources only; do not carry August values into September.
Fixed instruction
Report the period, closed count, eligible count and completion rate when both counts are known.
August run
18 / 24 = 75% completion for August.
September run
21 requests closed. Completion rate unavailable because the eligible count is missing.
Review gate
Ask for September's eligible count; do not reuse 24 or claim an improving rate.
The second run is the test: reusable structure must not become reused evidence.
Questions, adaptations & sourcesGo deeper when you need to
Is this the same as a saved style?
No. Style covers reusable visual and editorial preferences. Task instructions cover the reader question, required evidence, output and checks. You can use one style across many different task instructions.
Will my agent remember these instructions automatically?
Do not assume so. Store them in a location your chosen workflow supports and deliberately provide or retrieve them for the next task. Availability varies by client and setup.
Does this schedule the next report?
No. These instructions define how to create and review an edition. Scheduling, authorized data access and publishing operations are separate workflows, and the instruction sheet does not configure them.