@seanyao/roll 2026.514.5 → 2026.516.1

package/skills/roll-brief/SKILL.md

@@ -167,6 +167,7 @@ A simple heuristic — not a gate, just a signal for the human:
 ```bash
 git add docs/briefs/YYYY-MM-DD-NN.md
 git commit -m "docs: roll-brief YYYY-MM-DD-NN — {触发原因}"
+git push origin main
 ```
 
 - 触发原因来自调用上下文（Feature 完成 / 每日 / 手动 / `--feature` / `--since`），用一句话填入

package/skills/roll-review-pr/SKILL.md

@@ -0,0 +1,58 @@
+---
+name: roll-review-pr
+license: MIT
+allowed-tools: "Read"
+description: "Agent-agnostic PR review skill. Reviews a pull request diff and emits a structured 3-state verdict (APPROVE / REQUEST_CHANGES / UNCERTAIN). Used by `roll review-pr` and `_loop_pr_review_external`."
+---
+
+# PR Review
+
+> Follows the Architecture Constraints, Development Discipline, and Engineering
+> Common Sense defined in the project AGENTS.md.
+
+You are reviewing a pull request. Your job is to assess code quality,
+correctness, and adherence to project conventions.
+
+## Context
+
+**PR Title:** {{PR_TITLE}}
+
+**PR Body:**
+{{PR_BODY}}
+
+## Diff
+
+```diff
+{{PR_DIFF}}
+```
+
+## Review Instructions
+
+1. Read the diff carefully. Focus on:
+   - Correctness: logic errors, off-by-one, unhandled edge cases
+   - Security: injection, secrets exposure, unsafe operations
+   - Conventions: naming, structure, test coverage (as described in AGENTS.md)
+   - Scope: changes should match what the PR title/body claims
+
+2. Write your analysis in free text (2-10 sentences). Be specific — cite file
+   names and line numbers when pointing out issues.
+
+3. End your response with exactly ONE verdict footer on its own line:
+
+   - If the code is acceptable:
+     `<!--VERDICT:APPROVE-->`
+
+   - If changes are needed (cite the most important issue):
+     `<!--VERDICT:REQUEST_CHANGES:one-line reason-->`
+
+   - If you cannot confidently judge (e.g., missing context, domain-specific logic):
+     `<!--VERDICT:UNCERTAIN:one-line reason-->`
+
+## Rules
+
+- The verdict footer MUST appear on the last non-empty line of your response.
+- Choose exactly one verdict. Do not combine them.
+- REQUEST_CHANGES is for real issues — not style nitpicks or personal preferences.
+- When in doubt between APPROVE and UNCERTAIN, prefer UNCERTAIN.
+- If the PR body contains `[skip-ai-review]`, immediately output
+  `<!--VERDICT:APPROVE-->` with no analysis.

package/tools/roll-fetch/SKILL.md

@@ -1,182 +0,0 @@
----
-hidden: true
-name: roll-fetch
-description: Web page fetching and crawling for AI agents. Extract content from URLs for research, documentation, and competitive analysis.
----
-
-# Roll Fetch - Web Content Extraction
-
-Extract content from web pages for research and analysis.
-
-## When to Use
-
-- Product research (competitor analysis)
-- Technical documentation gathering
-- Code examples and best practices
-- Full site crawling for backup/analysis
-
-## Environment Setup
-
-Configure API keys per machine:
-
-```bash
-# Required for Tavily
-export TAVILY_API_KEY=tvly-dev-...
-
-# Optional for cloud browser fallback
-export BROWSER_USE_API_KEY=bu-...
-```
-
-Or create `.env` file in project root:
-```
-TAVILY_API_KEY=tvly-dev-...
-BROWSER_USE_API_KEY=bu-...
-```
-
-## Methods
-
-### 1. Tavily API (Recommended)
-
-Best quality extraction, requires `TAVILY_API_KEY`.
-
-```bash
-# Using Tavily CLI or API
-curl -X POST https://api.tavily.com/extract \
-  -H "Content-Type: application/json" \
-  -d '{
-    "urls": ["https://example.com"],
-    "api_key": "your_tavily_api_key"
-  }'
-```
-
-**Pros**: AI-optimized extraction, handles complex layouts
-**Cons**: Requires API key, rate limited
-
-### 2. LLM Native Fetch (Default)
-
-Use your built-in URL fetching capability directly.
-
-**When to use**: When Tavily is unavailable or for quick checks.
-
-**Note**: Most modern AI agents (Kimi, Codex, Claude) have native URL fetching. Use `FetchURL` tool or equivalent.
-
-### 3. Browser Automation (Fallback)
-
-Local browser automation for stubborn pages using **[browser-use](https://github.com/browser-use/browser-use)**.
-
-**How to Choose:**
-
-| If | Then Use | Why |
-|----|---------|-----|
-| `BROWSER_USE_API_KEY` in env | **Cloud** | Managed browsers, less setup |
-| No API key, but `browser-use` installed | **Local** | Free, no external dependency |
-| Neither | Skip to manual extraction | Tell user "Need browser automation setup" |
-
-**Option A: Local (Free, No API Key)**
-```python
-from browser_use import Agent, Browser, BrowserConfig
-import asyncio
-
-async def fetch_page(url):
-    # Pure local, no API key needed
-    browser = Browser(config=BrowserConfig(headless=True))
-    await browser.start()
-    page = await browser.get_current_page()
-    await page.goto(url)
-    content = await page.content()
-    await browser.stop()
-    return content
-
-# Run
-content = asyncio.run(fetch_page("https://example.com"))
-```
-
-**Option B: Cloud API**
-```python
-from browser_use import Agent
-
-agent = Agent(
-    task=f"Extract the main content from {url} and return as markdown",
-    llm="moonshot"  # or openai, anthropic
-)
-result = await agent.run()
-```
-
-**Setup** (Local):
-```bash
-pip install browser-use
-playwright install chromium
-```
-
-## Usage
-
-### CLI Usage (via smart-web-fetch.js)
-
-```bash
-# Auto mode (Tavily → Native → Browser)
-node smart-web-fetch.js fetch https://example.com
-
-# Explicit method
-node smart-web-fetch.js fetch https://example.com tavily
-node smart-web-fetch.js fetch https://example.com native
-node smart-web-fetch.js fetch https://example.com browser
-
-# Search
-node smart-web-fetch.js search "Python async" 5
-```
-
-### Programmatic Usage
-
-```javascript
-const { smartFetch, smartSearch } = require('./smart-web-fetch.js');
-
-// Fetch a page
-const result = await smartFetch('https://example.com');
-console.log(result.content);
-
-// Search
-const searchResult = await smartSearch('OpenAI GPT-5', 5);
-console.log(searchResult.results);
-```
-
-### Single Page Fetch
-
-```
-User: "Fetch https://docs.example.com/api"
-→ Use smart-web-fetch.js with auto mode
-→ Return clean markdown content
-```
-
-### Full Site Crawl
-
-```
-User: "Crawl https://docs.example.com"
-→ Use smart-web-fetch.js recursively
-→ Extract all internal links
-→ Recursively fetch up to max depth (default: 2)
-→ Save each page as separate markdown file
-```
-
-## Output Format
-
-Always return clean Markdown:
-- Extract main content only (remove nav, ads, footers)
-- Preserve code blocks and tables
-- Include source URL as header
-
-## Quality Check
-
-Validate extracted content:
-- Min length: 500 chars (reject if shorter)
-- Check for captcha/error messages
-- Verify main content structure (headings, paragraphs)
-
-## Examples
-
-| Task | Method | Command |
-|------|--------|---------|
-| Quick article | Auto | `node smart-web-fetch.js fetch https://blog.example.com` |
-| API docs | Tavily | `node smart-web-fetch.js fetch https://docs.example.com tavily` |
-| SPA site | Browser | `node smart-web-fetch.js fetch https://spa.example.com browser` |
-| Search | Tavily | `node smart-web-fetch.js search "Python async" 5` |
-| Fallback test | Native | `node smart-web-fetch.js fetch https://example.com native` |

package/tools/roll-fetch/package.json

@@ -1,15 +0,0 @@
-{
-  "name": "smart-web-fetch",
-  "version": "1.0.0",
-  "description": "Intelligent web fetching with automatic Tavily → Scrapling fallback",
-  "main": "smart-web-fetch.js",
-  "bin": {
-    "smart-web-fetch": "./smart-web-fetch.js"
-  },
-  "scripts": {
-    "test": "node smart-web-fetch.js fetch https://example.com"
-  },
-  "keywords": ["web-scraping", "tavily", "scrapling", "fallback"],
-  "author": "R0_lobster",
-  "license": "MIT"
-}