wuyang/research-blog
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat: publish Git collaboration and AI development research

Browse Source
This commit is contained in:
wuyang6
2026-08-10 02:49:12 +08:00
parent 8bd96c892a
commit 64ad40c9c6
15 changed files with 1339 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files
visuals/git-collaboration-ai-development/README.md
+104
View File
@@ -0,0 +1,104 @@
# Visual provenance
## 基本信息
- Renderer: OpenAI ImageGen，built-in mode
- Generated: 2026-08-10
- Article: `content/posts/git-collaboration-ai-development.md`
- Validation: 三张图片均已人工检查流程方向、中文标签、重复文字和水印；技术事实以正文为准。
## git-collaboration-flow.png
```text
Use case: infographic-diagram
Asset type: wide editorial infographic for a serious Chinese technical blog
Primary request: Create a clean horizontal infographic explaining the standard multi-person Git development workflow.
Scene/backdrop: warm off-white background, generous margins.
Style/medium: polished flat editorial infographic, vector-like shapes rendered as a raster image, restrained and professional, not playful, not futuristic.
Composition/framing: 16:9 landscape. One clear left-to-right flow with eight large nodes connected by simple arrows. A small role label is placed near each relevant stage. Strong visual hierarchy and ample whitespace.
Title text (verbatim): "多人协作开发流程"
Main node text, each verbatim and in this exact order:
"需求" → "方案" → "Issue" → "分支开发" → "PR / MR" → "CI" → "人工评审" → "合并发布"
Role labels, verbatim:
"产品负责人"
"技术负责人"
"开发者"
"工程系统"
"Reviewer"
Color palette: navy, muted teal, slate gray, a small amount of warm amber for decision points.
Typography: highly legible simplified Chinese sans serif; all text must be spelled exactly; large enough for a blog page.
Constraints: show only the listed labels; no extra prose; no company logos; no mascot; no decorative code; no watermark. Keep arrows unambiguous and avoid crossing lines.
Avoid: dense dashboard, 3D, gradients, neon colors, tiny text, English sentence fragments, duplicate nodes.
```
## ai-responsibility-lanes.png
```text
Use case: infographic-diagram
Asset type: wide editorial infographic for a serious Chinese technical blog
Primary request: Create a clear three-lane workflow showing the division of responsibilities after AI is integrated into software development.
Scene/backdrop: warm off-white background with generous whitespace.
Style/medium: polished flat editorial infographic, vector-like raster rendering, restrained, professional, calm.
Composition/framing: 16:9 landscape. Three horizontal swimlanes stacked vertically. Each lane has three large rounded cards connected left to right. Use subtle vertical alignment to show collaboration across lanes, but no tangled arrows.
Title text (verbatim): "AI 加入后的职责分工"
Lane 1 label (verbatim): "人"
Lane 1 cards, verbatim: "目标与验收" / "架构与风险" / "最终批准"
Lane 2 label (verbatim): "AI"
Lane 2 cards, verbatim: "调研与计划" / "编码与测试" / "失败分析"
Lane 3 label (verbatim): "工程系统"
Lane 3 cards, verbatim: "权限与隔离" / "CI 与门禁" / "记录与审计"
Color palette: human lane muted navy, AI lane muted teal, engineering system lane slate gray; small amber accents only at approval points.
Typography: highly legible simplified Chinese sans serif; all text must be spelled exactly and large enough for a blog page.
Constraints: show only the listed labels; no extra prose; no company logos; no robot mascot; no human faces; no watermark. Use simple abstract icons sparingly.
Avoid: dense dashboard, 3D, gradients, neon, tiny text, duplicate labels, illegible Chinese, decorative code snippets.
```
## ai-project-workflow-positions.png
```text
Use case: infographic-diagram
Asset type: wide editorial overview graphic for a serious Chinese technical blog
Primary request: Create a clean overview showing where five public AI development practices place AI in the software workflow. Do not rank products.
Scene/backdrop: warm off-white background, generous whitespace.
Style/medium: polished flat editorial infographic, vector-like raster rendering, restrained and professional.
Composition/framing: 16:9 landscape. A simple left-to-right development spine across the upper middle with four large stage labels. Below it, five distinct rounded cards positioned near the stage they influence. Cards may overlap adjacent stages visually through subtle connectors, but lines must remain simple.
Title text (verbatim): "AI 进入开发流程的位置"
Development stage labels, verbatim and in order: "任务与需求" → "实现" → "评审" → "仓库治理"
Five card titles and subtitles, all verbatim:
"OpenAI Codex" / "调研、实现与审查"
"Copilot coding agent" / "Issue 到 Draft PR"
"Claude Code Action" / "Issue / PR 触发"
"OpenClaw（项目实践）" / "规则、验证与自动审查"
"Agentic Workflows" / "事件驱动的持续自动化"
Color palette: navy, muted teal, slate gray, restrained amber accents.
Typography: highly legible simplified Chinese and English sans serif; all text must be spelled exactly; large enough for a blog.
Constraints: no ranking, no scores, no logos, no robot mascot, no company illustrations, no extra prose, no watermark. Make the distinction between a project practice, coding agents, an Action, and a workflow framework visually neutral.
Avoid: dense feature matrix, 3D, gradients, neon colors, tiny text, duplicate cards, illegible Chinese, crossed connectors.
```
## change-evidence-state.svg
- Asset type: deterministic Mermaid flow diagram.
- Renderer: `@mermaid-js/mermaid-cli`.
- Source: `change-evidence-state.mmd`.
- Purpose: explain the state, evidence and decision owner required as a change moves from problem discovery to production observation.
- Validation: manually checked node order, evidence labels, decision labels, arrow direction and desktop readability.
- Technical boundary: this is the article's general model, not a claim that every referenced repository implements identical states.
## test-selection-funnel.svg
- Asset type: deterministic Mermaid flow diagram.
- Renderer: `@mermaid-js/mermaid-cli`.
- Source: `test-selection-funnel.mmd`.
- Purpose: show how a test manifest is derived from Git Diff, directory rules, dependency graph, explicit risk mappings and historical regressions, with broad fallback when confidence is insufficient.
- Validation: manually checked selection order, fallback route, gate result and text wrapping.
- Technical boundary: the diagram combines verified public patterns with the article's generalized test-selection model; exact selectors differ by repository.
## agent-trust-boundary.svg
- Asset type: deterministic Mermaid security-boundary diagram.
- Renderer: `@mermaid-js/mermaid-cli`.
- Source: `agent-trust-boundary.mmd`.
- Purpose: separate untrusted input, read-only Agent execution, deterministic output validation, credentialed write execution and merge authorization.
- Validation: manually checked that the Agent never directly crosses into the credentialed write step and that CI/owner/human approval remains after the write request.
- Technical boundary: the diagram is a reference architecture synthesized from the cited official security designs, not the internal topology of any one vendor.