vision-analyze
Local vision analysis - screenshots, UI review, error detection via Ollama. 100% local. NOT visual regression.
- License 许可
- Apache-2.0
- 最近更新
- 2026-11-10
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- vision-analyze
- triggers
- [ "capture", "vision", "analyze-ui", "visual-review", "captura", "analizar-imagen" ]
- changelog
- docs/ciclos/cycle28-20260815.md
- description
- Local vision analysis - screenshots, UI review, error detection via Ollama. 100% local. NOT visual regression.
- token_budget
- 2664
03 SKILL.md 正文
When to Use
Local vision analysis — screenshots, UI review, error detection. Offline: if 127.0.0.1:11434 unreachable → degraded graceful, no crash (perf-offline-fallback.ps1).
Differentiation
vision-analyze = local LLM analysis via Ollama (127.0.0.1:11434) — semantic/feel review, NOT visual regression. Visual regression/pixel diffing → visual-testing (Playwright toHaveScreenshot).
Modes
ui layout/contrast/broken | error msgs+components | design spacing/typo/color | accessibility WCAG contrast/touch | performance CLS/images/loading
Security
Default: 100% local via 127.0.0.1:11434. Cloud allowed only if VISION_ANALYZE_OLLAMA_CLOUD=1 and OllamaApiKey set and screenshot sanitized (no PII) — see RUNBOOK.md and docs/mejoras/2026-08-27-ollama-cloud-investigation.md (G8 Option 1).
Hard Rules
- NEVER for visual regression/pixel diffing — that is
visual-testing(toHaveScreenshot); Ollama slow + non-deterministic - NEVER force
--model llava:7bwith <8GB free RAM — RAM-aware auto-select prevents OOM - Default: 100% local via 127.0.0.1:11434. Cloud allowed only if VISION_ANALYZE_OLLAMA_CLOUD=1 and OllamaApiKey set and screenshot sanitized (no PII)
- Offline-first fallback: if Ollama not reachable or allowlist blocks, degrade gracefully — no crash; verify server + models before any run
Output
VISION:<target>—<date> MODE:[ui|error|design|a11y|perf] MODEL:<name> ISSUES:<n> TOP:<issue> VERIFY:[screenshot|ollama]→<ok/fail>
Anti-Rationalization
| Rationalization | Red Flag | Verification |
|---|---|---|
| "Ollama caído → inventar análisis" | Offline sin check, hallucina | Verifica 127.0.0.1:11434 antes; si down → degraded graceful |
| "Usar para regression en vez de feel" | Pixel diff con Ollama no determinístico | Solo feel ui/error/design/a11y/perf; regression=visual-testing |
| "Sin fallback offline documentado" | Crash si Ollama bloqueado | Offline-first per perf-offline-fallback.ps1, no crash |
Red Flags
- Doing work without checking output format → STOP, re-read skill
- Second occurrence → force RED zone
Verification
- Output matches ## Output contract + file:line
- cross-ref-check.ps1 → SKILL.md OK
Cross-Refs: visual-testing | performance | accessibility | baseline-ui | ui-engine | web-quality-audit | seo | code-review-agent | self-improvement
Anti-Patterns
Use for pixel diff (visual-testing) · Force llava:7b <8GB (OOM) · Route unsanitized via external without VISION_ANALYZE_OLLAMA_CLOUD=1
docs/skills/vision-analyze/reference.md
04 同仓库更多 Skills
web-quality-audit
Comprehensive web audit: performance, a11y, SEO, responsive, animation, design tokens.
ui-engine
UI system — Grid/Flexbox/@layer/:has(), container queries, compositor-only animation, OKLCH tokens, component patterns
baseline-ui
Anti-slop UI — layout, typography, responsive, animation, tokens. Use for cleanup or polish.
skill-creator
Create new AI agent skills following Agent Skills spec -- auto-bootstrap unknown tasks, define structure, register in AGENTS.md
senior-engineer
Apply senior engineering competencies — system design, trade-off analysis, delegation, mentoring, and tech debt strategy
prompt-engineering
Design effective prompts using SPEARS framework — scope, principles, examples, assertions, refinements, with ReAct and multi-agent patterns