a11y-runtime-auditAudit a rendered web page for accessibility in a headless browser — a local dev server page or a Storybook story iframe URL — and report WCAG 2.2 findings, each with its success criterion, severity and a repro step: axe-core rules plus what static review cannot see (keyboard Tab walk with focus orde
dead-code-auditAudit a JavaScript/TypeScript repo for dead code with Knip, zero-install, and decide what is safe to remove: unused files, unused exports, unused exported types, unused dependencies and unlisted/undeclared dependencies, each categorized, with a before/after score (removable lines of code, unused-dep
front-reviewReview front-end code (React, TypeScript, CSS, Storybook, any stack): a pull request, a branch, a revision range, a path or the working diff. A script resolves the target and prints the review context: changed files grouped by rule axis, the repo's own REVIEW.md overlay read from the base branch, co
intakeGeneric front door for any new need, before designing anything: triage it (frontend/UI, backend analysis, bug, ticket, engineering management), look cheaply at what exists, interview the user in the main session, verify every load-bearing claim with evidence, reproduce or test when the result can ch
pr-guardianWatch one pushed branch in the background until its pull request is green, merged or blocked: a guardian agent rebases on the default branch, fixes what CI and the reviewers report, commits and pushes per the repo's configured policy, and reports with proof. One guardian per branch, a write lock so
pr-todoReport every open pull request a developer must act on in a GitHub organization — review requested, authored, or assigned — as a raw table with clickable links, merge status, the concrete reason each PR is stuck, and age. Use when someone asks which PRs are waiting on them, what reviews they owe, wh
sentry-triageRank the unresolved Sentry issues of a project by impact and decide what to fix first: a deterministic severity score, the trend, the innermost in-app frame, the failure origin (app / vendor / network / browser), the clusters sharing one root cause, then a complexity call per issue. Use when someone