BasicAgent

AI Prompt Examples (Governed Prompts for Reliable Outputs)

Copy/paste AI prompt examples for summarization, extraction, writing, and coding, with governance and system-prompt controls for consistent results.

People search “AI prompt examples” because they want prompts that work without guesswork. Here’s a small library of practical prompts you can copy, paste, and adapt—plus one reusable template that makes outputs more consistent.

Governance note

Prompt examples work best when paired with system controls and runtime monitoring. Use this with:

system prompting prompt tools ai governance

How to use these

Replace anything in {braces}.
Keep the output format section (that’s what makes results stable).
If the model rambles: tighten constraints (“Output JSON only”, “Max 120 words”, “Ask 1 question if unsure”).

A reusable prompt template (works for most tasks)

ROLE:
You are {role}. Be concise and correct.

TASK:
{what you want}

CONSTRAINTS:
- If information is missing, say what’s missing (don’t guess).
- Ask at most 1 clarifying question if needed.
- Use plain language.

OUTPUT:
{format: bullets / table / JSON / numbered steps}

Copy/paste prompt examples

1) Summarize a long document (decisions + action items)

Summarize the text below for a busy reader.

OUTPUT:
1) 5-bullet summary
2) Key decisions (if any)
3) Action items (owner → task → due date if present)
4) 3 important quotes (exact)

TEXT:
{paste}

2) Extract structured data (JSON you can feed to code)

Extract structured data from the text below.

SCHEMA (JSON):
{
  "entities": [{"name": "", "type": "", "evidence": ""}],
  "dates": [{"date": "", "context": ""}],
  "amounts": [{"amount": "", "currency": "", "context": ""}],
  "risks": [{"risk": "", "severity": "low|med|high", "evidence": ""}]
}

RULES:
- Output JSON only.
- Use empty arrays when none found.

TEXT:
{paste}

3) Rewrite copy to match a voice (without changing meaning)

Rewrite the text below in this voice:
- {voice adjectives, e.g. calm, confident, minimal}
- Avoid hype and buzzwords
- Keep meaning identical

OUTPUT:
- Version A (short)
- Version B (more detailed)

TEXT:
{paste}

4) Compare tools and recommend (with decision criteria)

Help me choose between these options: {A}, {B}, {C}.

CONTEXT:
{my context}

OUTPUT:
1) A decision table (criteria × option)
2) Your recommendation (1 paragraph)
3) “If you choose wrong” mitigation steps (3 bullets)

5) Turn messy notes into a clean SOP

Convert the notes below into a clear SOP someone can follow.

OUTPUT:
- Purpose (1 sentence)
- Preconditions
- Step-by-step procedure
- Common failures + fixes

NOTES:
{paste}

6) Debug a bug report (ask 1 question, then propose a fix)

You are a senior engineer. Diagnose the issue based on the report.

RULES:
- Ask exactly 1 clarifying question if needed.
- Then propose: (a) likely root cause, (b) fix steps, (c) how to verify.

BUG REPORT:
{paste}

7) Explain something to a non-technical user (no jargon)

Explain this to a non-technical user in plain language.

OUTPUT:
- 2-sentence explanation
- What it means for them
- What to do next (3 steps)

TOPIC:
{topic}

8) Generate supporting-page ideas (titles + intent + CTA)

Give me 12 supporting article ideas for the topic: {topic}.

RULES:
- Titles must be consumer-facing (no internal jargon).
- Each title must map to a real intent (how-to, comparison, troubleshooting).
- For each: primary keyword + 1-sentence promise + CTA target URL.