Cursor Team Kit
Cursor's official plugin bundle for internal developer workflows: CI, code review, shipping, local UI/CLI control, smoke tests, cleanup, PR review, and work summaries.
Current Source Snapshot
As of 2026-06-30, cursor/plugins is the public source of truth for the Team Kit. GitHub reports default branch main, HEAD 0452e08a314c03621ec5ac1324f1ad1dd824f1a4, 2,184 stars, and no repo-level license metadata; the cursor-team-kit/.cursor-plugin/plugin.json manifest says plugin version 1.2.0, category developer-tools, license MIT, and describes the bundle as Cursor internal workflows that do not require third-party service integrations. Source: GitHub cursor/plugins, 2026-06-30
The bundle contains 18 skills:
| Skill | Use |
|---|---|
check-compiler-errors |
compile/type-check failure reporting |
control-cli |
local CLI/TUI harness driving |
control-ui |
browser/CDP UI verification |
deslop |
AI-code cleanup |
fix-ci |
failing PR checks |
fix-merge-conflicts |
conflict resolution |
get-pr-comments |
review-comment intake |
loop-on-ci |
iterate until CI is green |
make-pr-easy-to-review |
reviewable PR shape |
new-branch-and-pr |
branch/PR setup |
pr-review-canvas |
PR review surface |
review-and-ship |
pre-ship review loop |
run-smoke-tests |
local smoke proof |
thermo-nuclear-code-quality-review |
deep simplicity review |
verify-this |
focused verification pass |
weekly-review |
work summary |
what-did-i-get-done |
activity summary |
workflow-from-chats |
workflow extraction from chats |
Routing Summary
Use Cursor Team Kit as the authoritative source page when evaluating Cursor's official workflow skills, importing a Cursor-specific skill, or comparing Cursor plugin structure against Kevin's generic skill runtime. Do not use it as the default route for code review, CI, QA, or tool hierarchy decisions unless the active harness is Cursor and the specific upstream skill has been inspected.
Routing Boundary
This is a Cursor-specific upstream bundle, not Kevin's generic agent brain. It belongs in the wiki as a source reference and candidate-import queue. Kevin's default review spine remains no-sus-code-doctor -> Security and Review Skills -> gstack-review, with ci-integrity-audit for green-washing checks and agent-browser / webapp-testing for runtime proof.
The useful general pattern is that serious agent teams package workflows as files: skills, agents, rules, and plugin manifests. That validates Kevin's runtime-index approach across Cursor, Codex, Claude Code, and .agents/skills, but the Team Kit's exact route names should not leak into generic routers unless the active harness is Cursor.
Field Signal
The high-signal bookmark is Parth Jadhav's field report: /thermo-nuclear-code-quality-review ran for about 30 minutes on a PR and materially improved code quality. The "67x" wording is a joke, not a benchmark. Preserve the operational lesson instead: deep skill passes can be long-running and still worth it when the target is simplicity, not syntax. Source: X/@ParthJadhav8, 2026-05-22
Skill Decision
Do not mass-import the 18 skills blindly. Import only when a skill fills a gap after comparing it against the existing MECE owners:
thermo-nuclear-code-quality-reviewoverlaps Cursor /thermo-nuclear-code-quality-review,no-sus-code-doctor, Security and Review Skills, andcode-taste.fix-ci,loop-on-ci, andciroutes overlapgh-fix-ci,ci-integrity-audit, andbabysit.control-ui/control-clioverlapagent-browser, browser verification skills, and local process-control skills.workflow-from-chatsoverlapsloop-me,learn-from-projects, and the no-one-off-work pipeline.
Promote a Team Kit skill only after reading its full upstream SKILL.md, preserving useful scripts/references, and adding a trigger eval that proves why Kevin's stack needs the new route.
Timeline
- 2026-06-30 | Created page from the one-by-one bookmark review; recorded plugin version
1.2.0, 18 skill names, the no-mass-import decision, and the generic-agent routing boundary. Source: X bookmark artifact audit, 2026-06-30 - 2026-05-22 | Parth Jadhav reported a roughly 30-minute
/thermo-nuclear-code-quality-reviewrun that improved a PR. Source: X/@ParthJadhav8, 2026-05-22