Prompt Library Entry Template
Use one entry per reusable prompt. Keep it concise enough to review before use. This record describes the tested configuration, evidence, and boundaries; it does not guarantee correct output.
Identity And Approval
- Name:
- Owner:
- Version:
- Status: candidate / approved / retired
- Approved use:
- Intended users:
- Approver, if required:
- Human-review rule:
Outcome And Use Boundary
- Intended outcome:
- Required inputs:
- Expected output:
- Non-approved uses:
- Sensitive-data boundary:
- Triggers for qualified review:
Prompt Template
[Paste the reusable prompt here.]
Placeholders
| Placeholder | Required | Source | Validation rule |
|---|---|---|---|
{{example}} | Yes | User input | Must be non-empty |
Worked Example
Input
[Representative input.]
Expected Output Shape
[Representative output or structure.]
Last-Tested Configuration
- Provider:
- Exact model identifier or snapshot:
- Configuration or exact configuration version:
- System instructions or immutable version/reference:
- Developer instructions or immutable version/reference; write “not used” if none:
- Tools, versions, and permissions:
- Retrieval sources, index, and version:
- Reasoning and generation settings:
Evaluation Evidence
- Test date:
- Representative cases:
- Comparison baseline:
- Acceptance criteria:
- Results:
- Material failures:
Limits And Fallback
- Known failure modes:
- Action when required inputs are missing or a check fails:
- Behaviour when the approved configuration is unavailable:
- Escalation owner or channel:
Change Record
- What changed:
- Reason for the change:
- Approver:
- Tests rerun and results:
- Trigger for the next reassessment:
Rerun the relevant tests when the prompt, model or snapshot, configuration, system or developer instructions, tools, retrieval sources, output contract, or policy changes, or after a material failure.