Prompt Quality

# Prompt Quality Rubric

Use this rubric when adding or reviewing prompts.

## Required

- The prompt has a specific target platform or assistant class.
- The intended mode is documented: text, voice, API, project, local agent, or other.
- The prompt includes a clear maintenance model for memory, state, or updates.
- The entry has measurable quality checks.
- Known limitations are stated plainly.
- Any persistence instruction explains what is manual, automatic, or simulated.

## Reject Or Revise

Revise prompts that:

- claim durable memory without explaining where memory is stored;
- require unavailable tools without fallback notes;
- include hidden instructions intended to bypass user control;
- include secrets, private data, or unredacted personal logs;
- ask the model to fabricate validation, citations, or results;
- add complexity without a testable benefit.

## Scoring

Use a 0-2 score for each area:

- Purpose clarity.
- Platform fit.
- Portability.
- Validation strength.
- Safety and privacy.
- Maintenance cost.

Prompts scoring below 8 out of 12 should remain draft unless there is a documented reason to publish early.