Prompt & Context Quality Evaluator
It does not produce a decorative prompt score. It evaluates each contract layer independently and ranks the highest-leverage fixes.
Primary output is portable as PDF, Markdown and machine-readable JSON.
The prompt should explicitly state what must be achieved.
→ Separate the objective from the task to execute.
The model should know which inputs it is allowed to use.
→ State available inputs and the data boundary explicitly.
Without constraints, the model tends to complete the scope on its own.
→ Define required and forbidden behavior as separate bullets.
A strong prompt defines the delivery shape, not only the task.
→ Add format, required fields, length and ordering to the output contract.
It should be possible to verify whether the output is good.
→ Add 3–7 verifiable acceptance criteria.
Behavior under missing information should be defined in advance.
→ Specify behavior for guesses, assumptions and missing critical data.
Context provenance and authority should be visible.
→ Rank sources as P0/P1/P2 and state authority explicitly.
Time-sensitive information needs an explicit freshness rule.
→ Add an as-of date or freshness verification rule.
Explicit exclusions reduce context noise.
→ State out-of-scope information and sources that must not be used.
0 estimated context tokens.
Promptfoo eval config
# Promptfoo eval scaffold generated by ahmetcanal.com
# Replace YOUR_PROVIDER/YOUR_MODEL before running.
description: "Prompt & Context Quality Evaluator"
prompts:
- |
[PROMPT REQUIRED]
providers:
- id: YOUR_PROVIDER:YOUR_MODEL
tests:
- description: "Basic output check"
assert:
- type: javascript
value: output.trim().length > 0
# Run locally with Promptfoo after choosing a provider/model.