michelbr84

Audit the full health of a GitHub repository. Triggers on /github-health commands and on natural-language requests like "check this repo", "is this PR safe to merge", "is this repo ready to release", or "audit security/CI/branches/Linear sync of this GitHub project". Routes to specialized sub-skills based on requested scope. Read-only by default; destructive actions require explicit user approval.

Audit Dependabot alerts and PRs in a GitHub repository: severity counts, affected packages, manifest and lockfile coverage, malware alerts, and version-update versus security-update configuration. Use this for vulnerability alerts and Dependabot PR backlogs; for the dependency tree itself use github-health-dependency-graph. Triggers on `/github-health dependabot <repo>` and requests like "what Dependabot alerts are open", "review vulnerable dependencies", or "is Dependabot configured". Read-only; never dismisses alerts or merges PRs.

Run an end-to-end health audit of a GitHub repository covering main branch, Actions, pull requests, branches, issues, security alerts, dependencies, documentation, releases, permissions, and Linear sync. Triggers on `/github-health full <repo>`, `/github-health standard <repo>`, `/github-health deep <repo>`, `/github-health quick <repo>` for a quick subset, and requests like "audit the whole repo" or "give me a complete health check". Read-only; destructive recommendations are approval-gated.

Apply score-improving changes to a GitHub repository based on the latest GitHub Health Report. Triggers on /github-health-improve commands and on natural-language requests like "raise this repo's health score", "apply the next improvement", "ratchet up the score". Reads the most recent saved report, picks the next highest-leverage improvement, optionally applies it on a non-main branch, opens a PR, and (when explicitly enabled) merges and re-audits. Approval-gated by default. Hard-stops on any red Action, conflict, secret detection, or destructive-operation requirement.

Run an umbrella security audit for a GitHub repository aggregating code scanning, Dependabot, secret scanning, malware alerts, repository security settings, SECURITY.md, and workflow permissions into one view. Use when the user wants the whole security picture; for a single dimension use child skills such as github-health-code-scanning, github-health-dependabot, or github-health-secret-scanning. Triggers on `/github-health security <repo>` and requests like "is this repo secure", "audit security posture", or "what is our security debt". Read-only; never dismisses alerts or changes settings without explicit approval.

Audit GitHub branch hygiene: merged branches, stale branches, orphan branches, branches behind main, branch naming risks, and branch cleanup candidates. Triggers on `/github-health branches <repo>` and requests like "review branches", "which branches can I delete", "branch cleanup", or "are there stale branches". Use this for branch-only cleanup; for cross-cutting cleanup (branches + PRs + issues + docs + security + Linear) use github-health-cleanup-plan instead. Read-only; never deletes branches without explicit approval.

Produce a safe, approval-gated multi-area cleanup plan for a GitHub repository covering branches, stale PRs, stale issues, documentation updates, security follow-ups, and Linear synchronization. Use this for cross-cutting cleanup; for branches-only cleanup use github-health-branches. Triggers on `/github-health cleanup-plan <repo>` and requests like "give me a cleanup plan" or "what should I tidy up across the repo". Read-only; every destructive step is explicitly approval-gated.

Audit CodeQL and other code scanning alerts in a GitHub repository: open alerts by severity, dismissed alerts with reasons, false-positive handling, and remediation tracking. Triggers on `/github-health code-scanning <repo>` and requests like "review CodeQL findings", "what SAST alerts are open", or "audit static-analysis results". Read-only; never dismisses alerts.

Gmail-first reporting skill suite. Analyzes inbox activity by day, week, month, year, or custom date range with flexible filters for unread messages, spam inclusion, labels, categories, senders, attachments, and response behavior. Generates structured reports in Markdown or PDF. Use when user says "gmail", "report", "inbox", "email", "unread", "spam", "senders", "labels", "response time", "trends", or any Gmail reporting request.

Inbox organization advisor. Detects unlabeled emails, clusters them by sender, domain, subject pattern, and content type, then suggests a complete label taxonomy with Gmail filter rules. Prescriptive counterpart to gmail-labels (which is descriptive). Supports --pdf and --pt (Portuguese).

Attachment analysis skill. Reports on attachment-heavy traffic, large-file patterns, file type distribution, and periods with high file volume.

Gmail category analysis. Breaks down email across Primary, Promotions, Social, Updates, and Forums tabs to measure inbox composition.

Custom date-range reports with mixed filters. Supports arbitrary start/end dates combined with any filter flags for fully flexible reporting.

Label distribution analysis. Breaks down email volume by Gmail labels to show how the inbox is organized and which labels receive the most traffic.

Resolves time period arguments (today, week, month, year, custom date range) into concrete start and end dates for Gmail data queries.

Markdown report generator. Takes analyzed data and produces a structured, well-formatted Markdown report suitable for Claude Code output, documentation, internal notes, and quick sharing.

Gera jogos em Pygame por gênero (ação, aventura, rpg, shooter, estratégia, puzzle, corrida, esportes, simulação, sandbox) usando templates e scaffold. Use quando o usuário pedir para criar/gerar um jogo Pygame, prototipar um gênero, criar um mini-game jogável, ou quando quiser validar automaticamente um jogo gerado (smoke test, screenshots).

Build a complete premium cinematic company website from an empty folder using Next.js, Tailwind CSS, and GSAP — full landing page, QA and build validation, secret scan, and a running localhost preview, with an optional Replicate asset pipeline. Use when the user asks for a full premium/Awwwards-style marketing or institutional site for a business.

Advances the Navecraft roadmap (TODO.md) by one step: reads TODO.md, picks the single best code-only next item, implements it end-to-end (code + tests), updates TODO.md honestly, and writes a complete report of what happened. Use this whenever the user wants to advance the phased roadmap tracked in TODO.md (not to edit an inline `# TODO` comment) — e.g. "fazer o proximo passo do TODO", "advance the roadmap", "advance the TODO", "pick and do the next thing", "do the next thing/item", "implementar o proximo item", "faz o proximo item", "continue the roadmap / TODO.md", or asks Claude to autonomously choose and ship the next pending task from TODO.md. Trigger on the slash command /proximo-passo, or on bare phrases like "proximo passo" / "next step" ONLY when the user clearly means advancing TODO.md (not a generic "what do I do next?" in an unrelated planning or debugging conversation).