UNPKG

paqad-ai

Version:

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

github.com/Eliyce/paqad-ai

Eliyce/paqad-ai

53 lines (38 loc) • 2.18 kB
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
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
--- name: ux-heuristic-evaluation description: 'Evaluates UI behavior against usability heuristics. Use when a meaningful UI change has known target flows and states, above all before final review or doc handoff.' metadata: model_tier: medium cacheable: 'true' output_format: markdown paqad-meta: skill.meta.yaml --- ## What It Does Evaluates changed UI behavior against a defined heuristic rubric so usability issues are reported as concrete user-impact findings instead of subjective design opinions. ## Use This When For any meaningful UI change once the target flow and states are known, especially before final review or documentation handoff. ## Inputs - Read the affected UI docs, flow notes, and current state inventory first. - Read `references/heuristic-rubric.md` before evaluating the interface. - Read the relevant stack-specific browser validation guide if behavior needs interactive confirmation. ## Procedure 1. Walk every heuristic in `assets/heuristics.txt` against the changed interface; never skip a category silently. 2. Tie each issue to a user task, state, or observed confusion point, never style preference. 3. Bucket findings into Blocking Issues vs Improvement Opportunities per `assets/output.template.md`. 4. Note any UI doc that must update because the heuristic reveals missing state documentation. 5. Validate with `scripts/lint-output.sh`. ## Output Contract - Return sections named `Blocking Issues` and `Improvement Opportunities`. - Each finding must state the affected screen or component, the heuristic violated, and the user impact. - If no issues are found, write `Blocking Issues: none` and `Improvement Opportunities: none`. ## Escalate / Stop Conditions - Ask when the expected user goal or task completion path is unclear. - Warn when the interface appears to conflict with accessibility or recovery expectations. - Do not file purely stylistic preferences as heuristic issues. ## Resources - `references/heuristic-rubric.md` - `scripts/lint-output.sh` - `assets/output.template.md` - `assets/heuristics.txt` - `runtime/capabilities/coding/stacks/flutter/references/tools/playwright.md` - `agents/openai.yaml`