vendor-docs-radar
Skill8 starsChanged 8 months agoMITView on GitHub
Scans official engineering blogs and documentation from Anthropic (Claude), Google (Gemini CLI), and OpenAI (Codex) for new posts, features, best practices, and breaking changes — then proposes concrete adoptions to the user's setup.
- Commits and pushes
In this repository5 other files it gives its agents.
- AGENTS.mdAGENTS.md
- agentic-radarSkill
- codex-agentSkill
- gemini-agentSkill
- reflect-and-learnSkill
The file
We split this skill into the rules it gives an agent
6 rules. About 0 are common, 5 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈5
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 30
When the user asks "what's new from Anthropic/Google/OpenAI"
- Distinctive: line 31
Before major config changes, to check for official guidance
- Distinctive: line 32
When a new version of Claude Code, Gemini CLI, or Codex is released
- Distinctive: line 262
Commit the report AND any auto-executed adoption changes together:
- Distinctive: line 339
When a vendor announces a breaking change, flag it as HIGH priority immediately
- Distinctive: line 341
If a vendor doc recommends a practice you're not following, that's always HIGH priority
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Vendor Docs Radar
Scans official engineering blogs and documentation from the three major AI coding assistant vendors — Anthropic (Claude Code), Google (Gemini CLI), and OpenAI (Codex) — for new posts, feature announcements, best practices, and breaking changes. Synthesizes findings and proposes concrete adoptions to the user's Claude Code setup.
When to Use
- Weekly scheduled runs (pair with agentic-radar for full coverage)
- When the user asks "what's new from Anthropic/Google/OpenAI"
- Before major config changes, to check for official guidance
- When a new version of Claude Code, Gemini CLI, or Codex is released
Output
- Dated report at
~/.claude/vendor-docs-reports/YYYY-MM-DD.md - Proposed adoption items (config changes, new features to enable, patterns to adopt)
- Updated
~/.claude/vendor-docs-reports/CHANGELOG.mdtracking what was found and when
Execution Instructions
Step 1: Setup
mkdir -p ~/.claude/vendor-docs-reportsSet DATE to today's date (YYYY-MM-DD).
Step 2: Launch 3 Parallel Scan Agents
Launch ALL three agents simultaneously as background agents with mode: bypassPermissions.
Each agent uses WebSearch and WebFetch to find and extract content.
Agent 1: Anthropic / Claude Official Scanner
Goal: Find new blog posts, engineering articles, documentation updates, and changelog entries from Anthropic related to Claude Code, Claude API, and Claude models.
Sources to scan (use WebSearch + WebFetch):
| Source | URL / Search Query |
| Anthropic Blog | site:anthropic.com/research OR site:anthropic.com/news — search for posts from last 30 days |
| Anthropic Engineering | site:anthropic.com/engineering |
| Claude Code Docs | site:docs.anthropic.com/en/docs/claude-code |
| Claude API Docs | site:docs.anthropic.com/en/docs/ — check for new/updated pages |
| Anthropic Cookbook | site:github.com/anthropics/anthropic-cookbook — recent commits/new notebooks |
| Claude Code GitHub | Fetch https://github.com/anthropics/claude-code/releases for new releases |
| Claude Code Changelog | Search "claude code" changelog OR "what's new" site:anthropic.com |
Search queries (run ALL):
anthropic blog(current year)anthropic engineering blog newclaude code new features(current year)docs.anthropic.com claude code updatesanthropic claude best practices(current year)claude code hooks MCP newanthropic claude model context protocol updatesclaude code CLI release notesanthropic prompt engineering guide updatesanthropic cookbook new notebooks(current year)
Extract for each finding:
- Title + URL
- Date published/updated
- Category:
blog|docs|changelog|cookbook|release - One-paragraph summary
- Relevance to our Claude Code setup (HIGH / MEDIUM / LOW)
- Specific actionable items (new flags, config options, best practices)
Return format: Markdown table of findings sorted by date (newest first), followed by a "Key Takeaways" section with actionable items.
Agent 2: Google Gemini CLI Scanner
Goal: Find new documentation, blog posts, release notes, and best practices for Gemini CLI and Gemini API that could improve Gemini agent integration.
Sources to scan (use WebSearch + WebFetch):
| Source | URL / Search Query |
| Gemini CLI GitHub | Fetch https://github.com/google-gemini/gemini-cli/releases for new releases |
| Gemini CLI Docs | site:github.com/google-gemini/gemini-cli — README, docs/, wiki |
| Google AI Blog | site:blog.google/technology/ai — Gemini related posts |
| Google Developers Blog | site:developers.googleblog.com gemini |
| Gemini API Docs | site:ai.google.dev/gemini-api/docs |
| Google Cloud AI | site:cloud.google.com/vertex-ai/generative-ai/docs — Gemini updates |
Search queries (run ALL):
gemini cli new features(current year)google gemini cli release notesgemini cli best practicesgemini api updates(current year)google gemini developer tools newgemini cli configuration optionsgemini pro new capabilities(current year)google ai studio gemini updatesgemini cli extensions pluginsgemini code assistance tools(current year)
Extract for each finding:
- Title + URL
- Date published/updated
- Category:
blog|docs|changelog|release|api-update - One-paragraph summary
- Relevance to gemini-agent skill (HIGH / MEDIUM / LOW)
- Specific actionable items (new CLI flags, model updates, API changes)
Return format: Same as Agent 1.
Agent 3: OpenAI Codex CLI Scanner
Goal: Find new documentation, blog posts, release notes, and best practices for OpenAI Codex CLI and OpenAI API that could improve Codex agent integration.
Sources to scan (use WebSearch + WebFetch):
| Source | URL / Search Query |
| Codex CLI GitHub | Fetch https://github.com/openai/codex/releases for new releases |
| Codex CLI Docs | site:github.com/openai/codex — README, docs/ |
| OpenAI Blog | site:openai.com/blog OR site:openai.com/index — Codex related |
| OpenAI API Docs | site:platform.openai.com/docs — new endpoints, models |
| OpenAI Cookbook | site:github.com/openai/openai-cookbook — recent additions |
| OpenAI Changelog | site:platform.openai.com/docs/changelog |
Search queries (run ALL):
openai codex cli new features(current year)openai codex cli release notesopenai codex best practicesopenai api updates(current year)openai developer tools newopenai codex configuration optionsopenai model new capabilities(current year)openai platform changelog(current year)openai codex cli extensions pluginsopenai coding assistant tools(current year)
Extract for each finding:
- Title + URL
- Date published/updated
- Category:
blog|docs|changelog|release|api-update - One-paragraph summary
- Relevance to codex-agent skill (HIGH / MEDIUM / LOW)
- Specific actionable items (new CLI flags, model updates, API changes)
Return format: Same as Agent 1.
Step 3: Synthesis (after all 3 agents complete)
Launch a synthesis agent that receives all three agent outputs and produces:
3a. Dated Report
Write to ~/.claude/vendor-docs-reports/YYYY-MM-DD.md:
---
date: YYYY-MM-DD
sources_scanned: [anthropic, google-gemini, openai]
---
# Vendor Docs Radar — YYYY-MM-DD
## Executive Summary
<!-- 3-5 bullet points of the most important findings across all vendors -->
## Anthropic / Claude
### New Posts & Docs
<!-- Table of findings from Agent 1 -->
### Actionable Items
<!-- Bulleted list of things to adopt/change -->
## Google / Gemini CLI
### New Posts & Docs
<!-- Table of findings from Agent 2 -->
### Actionable Items
## OpenAI / Codex
### New Posts & Docs
<!-- Table of findings from Agent 3 -->
### Actionable Items
## Cross-Vendor Trends
<!-- Patterns appearing across 2+ vendors — convergent features, shared best practices -->
## Proposed Adoptions
<!-- Specific changes to propose to the user's Claude Code setup -->
| Priority | Source | Change | Rationale |
|----------|--------|--------|-----------|
| HIGH | ... | ... | ... |
| MEDIUM | ... | ... | ... |
## Sources
<!-- All URLs referenced in this report -->3b. Update Changelog
Append to ~/.claude/vendor-docs-reports/CHANGELOG.md:
## YYYY-MM-DD
- Scanned: Anthropic, Google Gemini, OpenAI
- Found: N new items (X high-relevance)
- Key: <one-line summary of most important finding>Step 4: Execute and Propose Adoptions
Classify each adoption item by safety:
Auto-execute (P0/P1, safe): Apply immediately without asking.
- Model name/version updates in skill configs
- New config additions (flags, settings) that are additive
- Language softening or clarification in docs
- New references, URLs, or source links
- Adding new search queries or scan sources to this skill
Flag for user review (HIGH-IMPACT): Present as proposals, do NOT auto-apply.
- Changes to execution style (Section 1 of CLAUDE.md)
- Permission or security-related changes
- Removing or replacing existing functionality
- Breaking changes to APIs or CLIs we depend on
For each auto-executed item, log what was changed and why (with source URL). For each flagged item, present: specific file + diff, source URL, risk level.
Step 5: Commit All Changes
Commit the report AND any auto-executed adoption changes together:
cd ~/.claude && git add -A && git commit -m "vendor-docs-radar: scan YYYY-MM-DD + auto-adopt P0/P1" && git pushAutomation
To run weekly, create a launchd plist at ~/Library/LaunchAgents/com.claude.vendor-docs-radar.plist:
<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
<plist version="1.0">
<dict>
<key>Label</key>
<string>com.claude.vendor-docs-radar</string>
<key>ProgramArguments</key>
<array>
<string>/bin/bash</string>
<string>-l</string>
<string>-c</string>
<string>~/.claude/scripts/vendor-docs-radar.sh</string>
</array>
<key>StartCalendarInterval</key>
<dict>
<key>Weekday</key>
<integer>5</integer>
<key>Hour</key>
<integer>3</integer>
<key>Minute</key>
<integer>0</integer>
</dict>
<key>StandardOutPath</key>
<string>/tmp/vendor-docs-radar.log</string>
<key>StandardErrorPath</key>
<string>/tmp/vendor-docs-radar.err</string>
</dict>
</plist>Script: ~/.claude/scripts/vendor-docs-radar.sh
#!/bin/bash
cd ~/.claude
claude -p "Run the vendor-docs-radar skill. Today is $(date +%Y-%m-%d)."For Linux, use cron instead:
0 3 * * 5 ~/.claude/scripts/vendor-docs-radar.shDifferences from agentic-radar
| Aspect | vendor-docs-radar | agentic-radar |
| Focus | Official vendor docs & blogs | Community repos & discussions |
| Sources | anthropic.com, ai.google.dev, openai.com, platform.openai.com | GitHub trending, HN, Reddit |
| Signal type | Authoritative best practices, new features, breaking changes | Emerging tools, community patterns |
| Agents | 3 (one per vendor) + synthesis | 5 (GitHub, community, registry, deep-dive, audit) |
| Output | Dated report + adoption proposals | Report + registry updates + CLAUDE.md diffs |
| Cadence | Weekly (Fridays) | Weekly (Mondays) |
Use both together for full coverage: vendor-docs-radar catches official announcements, agentic-radar catches community innovations.
Notes
- Prioritize Anthropic findings — that's the primary platform
- Cross-vendor trends (features appearing in 2+ platforms) are especially valuable signals
- When a vendor announces a breaking change, flag it as HIGH priority immediately
- Compare findings week-over-week to track velocity of changes
- If a vendor doc recommends a practice you're not following, that's always HIGH priority
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

