Lockfile Watch for daily.dev
A daily.dev plugin that checks posts about compromised and malicious packages against your lockfiles, on your machine. It sends the alert to your phone only when you are, or might be, affected.
Updated
Lockfile Watch is a plugin in the daily.dev plugin marketplace. Your agent searches daily.dev for posts about compromised, malicious and malware packages. It matches the packages those posts name against the lockfiles in your repositories, on your computer. If you use none of them, it says nothing.
With Pushary connected, the alert reaches your phone, with what to do next. Lockfile Watch only reads and reports. It never changes your repositories.
Install
claude plugin marketplace add dailydotdev/marketplace
claude plugin install lockfile-watch@daily-dev-communityCodex reads the same marketplace with codex plugin marketplace add dailydotdev/marketplace. In Cursor (Teams and Enterprise), import https://github.com/dailydotdev/marketplace under Plugins & MCPs, Team Marketplaces.
daily.dev reviews every plugin before it lists it. Until the review is done, these commands do not find Lockfile Watch yet.
Then connect daily.dev to your agent. Add the daily.dev MCP server (https://api.daily.dev/mcp), or put a personal access token from daily.dev, Settings, API Access, in DAILY_DEV_TOKEN.
What it tells you
| Answer | What it means | Phone |
|---|---|---|
| Affected | Your lockfile has a bad version, or the malicious dependency the attack added | Yes |
| At risk | Your locked version is fine, but your version range allows a bad one | Yes |
| Unknown | You use the package, and there is no advisory yet or no one has published the bad versions | Yes |
| Not affected | An advisory covers the attack, and nothing you lock or allow is bad | No, terminal only |
Each finding alerts once, and again only if its answer changes. Only the GitHub and OSV advisory databases can clear a version. Write-ups and daily.dev collections can only add bad versions, because anyone can write a page or share a link.
Each alert says what to do. If you are affected and the bad version was ever installed, its install script may already have run, so start by rotating the secrets that machine could reach.
Get the alert on your phone
Run setup on the computer where your agent runs, then open the link on your phone:
npx pushary@latest setupSee the quickstart for what you need. Pushary is a paid service.
Run it every day
Run it on the computer where your repositories live: a launchd or cron job that starts your agent in a folder of repositories, or a Codex automation. launchd runs a missed job when the Mac wakes, and cron does not. Allow the tools it uses in advance, because nobody is there to approve them: the daily.dev tools, Pushary's send_notification, web fetch, rg or grep, git rev-parse, npm view, curl for the GitHub and OSV advisory APIs only, and writing its seen file inside each repository's .git folder.
One run checks every repository you list and makes three daily.dev searches. A free daily.dev account has 200 requests a month, so a daily run uses about 90.
What leaves your computer
- daily.dev gets three search words (compromised, malicious, malware) and post ids. Never your dependency list.
- The package registry, the GitHub and OSV advisory databases, and the write-ups get the name and locked version of a package that matched, to check versions and publish dates.
- Pushary gets package names, versions and repository folder names inside an alert, only when something is Affected, At risk or Unknown.
Your agent's model provider sees the files your agent reads, as with any task.