aiproof-cli 0.1.2

CLI for aiproof, the static analyzer for AI prompts (ESLint for prompts).
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28

# AIP017: system-message-mismatch

## Description

Detects Anthropic-style system prompts (opening with "You are...") when targeting Gemini models. Gemini handles system messages with different semantics than Anthropic's Claude, and direct porting may not work as intended.

## Severity

Info

## Applies When

This rule only fires when `target_models` is configured to include models with names starting with `gemini` (case-insensitive). The rule only checks `system` role prompts. If no target models are specified, this rule produces no diagnostics.

## Example

```
role: system
model: gemini-1.5-pro
prompt: |
  You are a helpful assistant that answers questions accurately.
```

Violation: Gemini's system message handling differs from Claude's; this pattern may not achieve the desired behavior.

## Remediation

Consult Gemini's documentation on system instructions and restructure your prompt to align with Gemini's conventions. Consider using explicit instructions instead of role-play framing.