REPO PACKAGE
getsentry warden
OFFICIAL11skills
8ready to use
LABSCO SUMMARY
Five of the eleven are the actual review content: code-review hunts real correctness bugs, security-review runs OWASP-style checks for injection, XSS, and unsafe crypto, architecture-review is a staff-level codebase health pass for monolithic modules and coverage gaps, agent-prompt is Warden's own guide to writing skills like these, and skill-writer creates or edits skills against the Agent Skills specification. warden and warden-sweep are the entry points that actually run a review — against the current branch's changes, or a full-repository sweep that verifies findings through deep tracing and opens draft pull requests — and both need WARDEN_OPENAI_API_KEY or WARDEN_ANTHROPIC_API_KEY set, since the reviewing happens in Warden's own CLI and GitHub Action, not in the skill file alone.
This is for a team that wants Sentry's own review checklists wired into a pull request flow: npx @sentry/warden init, then add security-review or add code-review, then either run the CLI before committing or let the GitHub Action comment inline on every PR and report to Checks. If you already run a different AI review tool, there's nothing here you couldn't get by reading Warden's product docs directly — the skills are the same instructions the CLI itself runs.
READ THE FULL ANALYSIS
Three of the eleven are test fixtures, not something to install. test-skill, test-skill-a, and test-skill-b exist for Warden's own integration tests, and their descriptions say exactly that — "A test skill for integration tests" — with nothing an agent could act on. Nothing in the package marks them differently from the five real reference skills, so they arrive alongside them.
Five more have already gone dead. find-bugs, find-warden-bugs, notseer, vercel-react-best-practices, and warden-skill all once existed in this repository and no longer do; none carry a description in what we could read, so we can't say what they covered or why Sentry retired them, only that today's eleven replaced them.
ALSO IN THIS PACKAGE
PLUGINS
1 itemPLUGIN
warden
Run Warden and query Warden Service
MCP SERVERS
1 itemMCP SERVER
sentry
Official Sentry MCP server for investigating issues, error reports, traces, and performance monitoring data from AI coding agents.
http · https://mcp.sentry.dev/mcp/sentry/wardenWHAT'S INSIDE
11 showing · 11 totalReady to use
Nothing else to set up — install it and go.
11 of 11
SETUP
Showing all 11 skills
agent-prompt
Advice on how to word the instructions you hand an AI assistant, collected into eight short notes — layout, phrasing, the shape of the answer you want back, the mistakes people keep repeating — with the right one pulled out for the question you asked.
architecture-review
A senior engineer's read of an entire codebase — not the part you just changed — saying what is quietly rotting, exactly where, and what to do about it, worst first.
code-review
Somebody reads the code you just changed while actively trying to break it, and stays silent unless they can point at the input that makes it go wrong — so what comes back is real bugs, not opinions.
security-review
Reads your code changes the way someone trying to break in would: it follows anything an outsider could type into the system and reports every place that value ends up somewhere it should never reach.
skill-writer
A skill is a self-contained set of directions that teaches an AI assistant one job; this is the agreed procedure for building one — read the source material first, then write it, then file it where the tool will find it and check it, with anything too thin sent back.
test-skill
test-skill-a
test-skill-b
testing-guidelines
What this project expects from its tests: outside services are mocked, fixtures come from real responses with private data scrubbed out, every bug fix ships with a test that would have caught it, and each command, route, or exported function has one of its own.
warden
The last check you run before handing your work in: it reads whatever you have just changed and sorts what it finds into must-fix, worth-a-look and minor.
warden-sweep
Point it at a whole codebase and walk away: it works through every file, throws out the findings it cannot stand behind, and comes back with a counted report and a pile of proposed fixes waiting for someone to approve.
No skill in this package matches that.
HOW TO GET IT
WHOLE PACKAGEInstall the whole package (11 skills)
npx skills add getsentry/wardenSINGLE SKILLOr install a single skill
npx skills add getsentry/warden --skill <name> --full-depthPick the skill name from the Skills tab — each entry there installs independently.
PLUGIN MARKETPLACEAdd the marketplace, then install the plugin
/plugin marketplace add getsentry/warden/plugin install warden@wardenTyped inside the agent's own prompt, not in a terminal. The marketplace is called warden, which is the part after the @.