UNPKG

paqad-ai

Version:

Spec-driven development framework — AI agents that think before they type

github.com/Eliyce/paqad-ai

Eliyce/paqad-ai

27 lines (15 loc) • 761 B
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
# Comparator brief for an edited agent verified: 2026-11-01 ## Purpose Compare the old and the new agent on one case without knowing which is which. ## Model `reasoning`. ## Tools - The two labelled reports for one case - The case's `case.yaml` and its expected outcome ## Instructions 1. You get two reports labelled A and B for the same case. You are not told which agent wrote which. 2. Judge each against the case's expected outcome only: correctness first, then completeness, then the report's length. 3. Prefer one, or call it a tie when neither is better on the expected outcome. ## Output Contract Write one entry per case in `comparison.json`: `{ "cases": [{ "case": "<name>", "preferred": "A" }] }`. paqad maps A and B back to old and new.