Guide · Agent skill
How to use the Canary Watch agent skill after deploys
Updated · written with AI from the project's own documentation and checked against our catalogue data
Canary Watch is an agent skill that tells an AI agent to check a deployed URL for regressions after you deploy a site or app, merge risky code, or upgrade dependencies. This guide is for developers and small teams who ship to production and want a quick, repeatable post-deploy check. By the end you will know what the skill checks, how to install it, how to trigger it with a request, and what it needs to run.
Canary Watch (デプロイ後監視)affaan-m/ECC · MIT Download the skill Quick start
- 1Download the ZIP. Download the canary-watch ZIP from this page and unzip it.
- 2Place the skill folder. Put the canary-watch folder in ~/.claude/skills/ for all projects, or in your project's .claude/skills/ folder for one project.
- 3Prepare a baseline and optional alerts. Decide which URL to watch and, if you want alerts, set up a desktop notification or a Slack or Discord webhook URL. Make sure you have a browser or headless browser tool for console and Core Web Vitals checks.
- 4Run a quick check. Ask the agent or run /canary-watch with your URL to get a single-pass report.
- 5Start a watch for a launch window. Use --interval and --duration to repeat checks during the window you care about.
- 6Review the report. Read the status and diffs from baseline, and act on any critical items.
What the Canary Watch skill makes the agent do
The skill directs the agent to monitor a deployed URL and look for regressions compared with what was there before. It runs a loop until it is stopped or its monitoring window ends, and it checks six things on each pass: HTTP status, new console errors, failed or 5xx network responses, Core Web Vitals (LCP, CLS and INP) against a baseline, whether key page elements such as the h1, nav, footer and CTA are still present, and whether important API endpoints respond within SLA.
The skill reports status and differences from baseline. It can alert on critical issues through desktop notifications, an optional Slack or Discord webhook, and a log file at ~/.claude/canary-watch.log.
- Quick check: a single pass that reports results
- Continuous watch: repeats at an interval for a set duration
- Compare mode: checks staging against production
Install the canary-watch skill
Download the ZIP from this page and unzip it. To make the skill available to all your projects, place the canary-watch folder in ~/.claude/skills/. To limit it to one project, place it in that project's .claude/skills/ folder.
You can also upload the skill in the Claude apps. Go to Settings, then Capabilities, then Skills, and upload it there.
- Personal install: ~/.claude/skills/
- Project install: .claude/skills/ inside your project
- Claude apps: Settings → Capabilities → Skills
Run a quick check or a continuous watch
The default mode is a quick check: one pass that reports the results. Pass the URL after the skill command.
For a watch during a launch window, add an interval and a duration. The source shows an interval of 5 minutes and a duration of 2 hours as examples. To compare staging with production, use compare mode and give both URLs.
/canary-watch https://myapp.com
/canary-watch https://myapp.com --interval 5m --duration 2h
/canary-watch --compare https://staging.myapp.com https://myapp.comTrigger the skill with natural requests
You can invoke the skill with the slash command above. The skill's description also matches requests about monitoring a URL after a deployment, merge or dependency upgrade, so you can phrase it in plain language, for example asking the agent to watch the live site for regressions for the next hour after a release.
Name the URL and say why you are checking, such as after a deploy or to confirm a fix. The more specific the request, the easier it is for the agent to pick the right mode.
- "Watch https://myapp.com for regressions for the next hour after this deploy."
- "Compare staging with production after I merged the checkout change."
- "Check that the fix for the login error is live on myapp.com."
Understand the alert levels
The skill sorts findings into three levels. Critical items trigger an immediate alert. These are a non-200 HTTP status, more than 5 new console errors, LCP above 4 seconds, or an API endpoint returning 5xx. Warnings appear in the report and include LCP rising more than 500ms over baseline, CLS above 0.1, new console warnings, and response times above twice the baseline. Info items are only logged, such as minor performance variation or new network requests like an added third-party script.
Read the report
Each run produces a markdown report with a status section and a diffs-from-baseline section. The status section lists passing checks, and the diff section shows changes such as CLS, response time and network request count, with notes on what to investigate.
- Check the HTTP status and critical errors first
- Review diffs for new third-party requests
- Keep the log file to compare runs over time
What people use it for
- Checking a production site a few minutes after a deploy for HTTP errors and missing navigation
- Watching a launch window for Core Web Vitals regressions such as LCP increases
- Comparing a staging URL with production before promoting a change
- Confirming that a bug fix is live on the public URL
- Monitoring a site after a dependency upgrade for new console errors or failed API calls
Frequently asked questions
How do I install the canary-watch agent skill?
Download the ZIP from this page and unzip it. Place the canary-watch folder in ~/.claude/skills/ for all projects, or in a project's .claude/skills/ folder for one project. You can also upload it in the Claude apps under Settings, Capabilities, then Skills.
What does canary-watch need to run?
It needs browser or headless browser tooling to measure console errors and Core Web Vitals; the skill file does not name a specific tool. Slack or Discord webhook URLs and desktop notification support are optional, and it needs write access to ~/.claude/canary-watch.log for logging. It runs scripts and uses the network.
Is canary-watch free and what is its licence?
The item is listed under the MIT licence. The source does not describe pricing, so check the licence text and project page for details on any costs.
Can canary-watch replace my test suite?
No. It is a monitoring pass, not a test framework. For full functional or UI tests, use dedicated QA tools; the skill itself suggests pairing with browser-qa for complete UI testing.
Is it safe to run canary-watch against any website?
Only run it against sites you own or have permission to hit repeatedly. Continuous polling may be unwelcome or look like abuse to the site owner, and the skill makes network requests on each pass.
How is canary-watch different from benchmark or browser-qa?
The source describes canary-watch as a post-deploy regression monitor. It can be paired with /benchmark for performance comparisons and with /browser-qa for full UI testing, so each tool covers a different part of the release check.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Canary Watch suits developers and small teams who ship to production and want a repeatable check after each deploy or launch window, provided they have a browser tool set up and permission to monitor the site. If you need full functional or UI test coverage, or you are checking a site you do not control, use a dedicated QA tool or a monitoring service instead.
Download the skill