Triggers
- "go deploy"
- "deploy platform"
- "prepare to deploy"
On trigger, execute one action: open the deployment platform link. No deployment execution, no status queries, no build operations.
Phase 1: Read docs/deploy.md
- Look for
docs/deploy.mdin the project root - Exists → proceed to Phase 3 (parse links)
- Does not exist → proceed to Phase 2 (scan project)
Phase 2: Scan Project
docs/deploy.md doesn't exist — the project hasn't configured deployment info yet. Scan existing docs to auto-extract and generate config, reducing manual input burden.
Scan these sources in priority order (higher-priority sources are typically maintained by project owners and more reliable):
| Priority | Source | Scan Logic |
|---|---|---|
| 1 | AGENTS.md |
Find sections with "部署" "deploy" "发布" keywords, extract URLs |
| 2 | README.md |
Same as above |
| 3 | scripts/ directory |
Find scripts with deploy/open/publish in filename, read URL variables |
| 4 | CI/CD config | .github/workflows/*.yml, .gitlab-ci.yml, Jenkinsfile |
| 5 | package.json |
scripts field with deploy/publish/release commands |
Link filtering: Project docs often contain many irrelevant URLs (dependency sites, reference docs, etc.). Only extract URLs that look like deployment consoles (containing deploy/console/cluster/release keywords, or known platform domains). Deduplicate and annotate each with its source.
Scan Result Handling
- Candidate links found → list all candidates (with source files), ask user to confirm which are actual deployment links (keyword matching has false positives — user confirmation avoids writing wrong info) → after confirmation, generate
docs/deploy.md(see format below) → proceed to Phase 3 - No candidate links found → output this prompt:
No deployment-related information found in the project.
Please provide the following and I'll create docs/deploy.md:
1. Deployment link URL (required)
2. Deployment platform name (optional, e.g. "Vercel", "Internal Deploy Platform")
3. Brief description (optional)
If the project has multiple deployment links, list them separately.
After user provides info, generate docs/deploy.md and proceed to Phase 3.
Phase 3: Parse Links
Extract deployment links from docs/deploy.md:
- Find all
##level-2 headings (each heading = one deployment target) - Under each heading, find lines starting with
**Link**:or**链接**:or**URL**:and extract the URL - Optionally read
**Platform**:/**平台**:and**Description**:/**说明**:lines - Aggregate into a link list
Error handling: docs/deploy.md exists but no link line found → prompt that format doesn't match convention, proceed to Phase 2 (scan). URL not starting with http:// or https:// → skip and warn.
Phase 4: Select Link
- Only 1 link → proceed directly to Phase 5
- Multiple links → list all, ask user to select (ask every time, don't remember previous choice):
Detected the following deployment links:
1. <first ## heading>
2. <second ## heading>
Enter the number of the link to open:
Phase 5: Open Link
Execute the OS-appropriate command:
# macOS
open "<URL>"
# Linux
xdg-open "<URL>"
# Windows (Git Bash / MSYS2 / Cygwin)
cmd.exe /c start "" "<URL>"
Fallback: If browser can't open or command fails → output the URL directly, prompt user to copy manually. After opening, state: "Opened: <link name>".
docs/deploy.md Format Convention
Each deployment target is a ## level-2 heading. Under it, a list with **Link**: line for URL (required); **Platform**: and **Description**: are optional. Create the docs/ directory if it doesn't exist when generating.
Example:
# Deployment
## Production
- **Platform**: Internal Deploy Platform
- **Link**: https://deploy.example.com/console/my-project
- **Description**: Main deployment entry — view cluster status, release new versions
## Staging
- **Link**: https://pre-deploy.example.com/my-project