getsentry sentry-javascript
OFFICIALLABSCO SUMMARY
Five cover releasing and packaging the SDK: release walks through publishing a new version and updating the changelog, add-cdn-bundle and bump-size-limit manage the browser package's CDN bundle and its enforced size budget, and upgrade-dep and vendor-otel handle bringing in or inlining a dependency such as OpenTelemetry. Another five are quality gates for a change before it merges: e2e and write-tests run and author the SDK's Playwright and Vitest suites, fix-issue and fix-security-vulnerability attempt a scoped fix for a GitHub issue or a Dependabot alert, and triage-issue researches an issue and writes up a recommendation without necessarily fixing it.
The remaining seven step outside those two jobs. add-ai-integration adds a new AI-provider instrumentation to the SDK itself — extending what it monitors, rather than building, testing, or releasing it. skill-creator and skill-scanner build and audit the repository's own agent skills, dotagents manages which of them are installed, and track-framework-updates produces a weekly digest of upstream framework activity for the team. linear-project-status and linear-project-update read and post to a Linear project — Sentry's own internal project-tracking workspace, not anything a user of the SDK has access to.
READ THE FULL ANALYSIS
The readme says nothing about any of this. getsentry/sentry-javascript's actual readme lists the SDK's supported platforms and links to docs, Discord, and the bug bounty program — the usual page for a product, with no mention of skills or an agent anywhere in it. We read these seventeen directly out of the repository.
What it costs to use one. Twelve of the seventeen need a local tool already in place — mostly the SDK's own build, lint, and test scripts. Two, the Linear pair, need a live MCP connection to Sentry's Linear workspace, which an outside contributor won't have. Two are ready with no setup (add-ai-integration and write-tests), and one, triage-issue, is unverified in our data.
WHAT'S INSIDE
17 showing · 17 totalNothing else to set up — install it and go.
add-ai-integration
Adds one more AI service to the list Sentry's JavaScript monitoring library can track, so that an app's calls to it get recorded the same way the existing ones are.
add-cdn-bundle
A ready-made copy of Sentry's browser code carrying exactly the features you pick, so that a website can load it in one line instead of building its own.
bump-size-limit
When a code change pushes one of the project's built files past the size it is allowed to be, this measures how big it really got and lifts that one limit by just enough.
dotagents
A package manager for the add-on skills an AI coding assistant uses: one file lists what the project needs, and one command fetches them and puts them where Claude Code, Cursor and the rest will look.
e2e
Installs your changed Sentry code into a full sample application and runs that application's own tests, so you learn whether the change works in real use and not only in isolation.
fix-issue
Takes one bug report filed against Sentry's JavaScript project and tries to close it unsupervised, under a hard rule that the change stays small and safe or is not made at all.
fix-security-vulnerability
Works through GitHub's security warnings about the outside code a project depends on, deciding for each one whether to upgrade the package or to wave the warning off with a written reason.
linear-project-status
A weekly once-over of a project tracked in Linear, for the person running it: is it slipping, what is blocking it, and what to put right this week.
linear-project-update
Drafts the progress update for a project you run in Linear, in the voice you normally write in, and posts nothing until you have read it and said go.
release
Prepares the next version of Sentry's JavaScript library for publishing: settles the version number, writes up everything that changed since the last one, and leaves it for a colleague to approve.
skill-creator
Walks you from a rough idea for a new Claude skill to a finished one, then settles whether it actually helps by running the same test requests with the skill and without it so you can compare.
skill-scanner
Picks apart a skill written by someone else, looking for hidden instructions, stolen data or quiet edits to your assistant's settings, and ends on install, install carefully, or don't.
track-framework-updates
A weekly roundup of what moved in the outside frameworks Sentry's JavaScript library plugs into, marked up with whether each change is something Sentry now has to deal with.
triage-issue
Reads a newly filed bug report and digs through the code to work out what is probably causing it, how urgent it is, and which files someone would have to look at.
upgrade-dep
Moves one outside package the project depends on up to a newer version and checks that everything still builds, stopping short of the version jumps that are known to break things here.
vendor-otel
Copies the source code of an outside tracing package straight into Sentry's own codebase, so the project stops installing that package and owns and maintains the code itself.
write-tests
The Sentry team's rules for writing tests: a few sharp ones that each pin down a single behaviour, in place of a pile of shallow ones that pass no matter what breaks.
HOW TO GET IT
npx skills add getsentry/sentry-javascriptnpx skills add getsentry/sentry-javascript --skill <name> --full-depthPick the skill name from the Skills tab — each entry there installs independently.