MichaelVessia

Run two heterogeneous fresh-context Herdr reviewers against both code quality and a governing plan, then synthesize evidence-backed feedback. Use when the user asks for a dual or adversarial review, code-versus-plan verification, or independent Claude and Pi reviewers before deciding whether a PR needs changes.

Control herdr from Codex. Use when the user says "use Herder" or asks to inspect, focus, prompt, spawn, or manage herdr workspaces, tabs, panes, or agents from a Codex session, without deciding whether they are inside or outside herdr.

Use when the user explicitly asks Codex Desktop or another shell outside Herdr to create, inspect, prompt, follow up with, or monitor Herdr agent tabs, or when the `codex-herdr` router skill hands off external control.

Fan out a user-provided task list into one reviewed PR per task through Herdr, then shepherd the PR fleet through CI and review without merging. Use when the user asks to fan out PRs, parallelize backlog items into separate PRs, or autonomously drive multiple agent-owned PRs to green.

Inspect and import content into my self-hosted AutoCaliWeb library. Use for catalog status, search, recent imports, shelves, metadata, or importing an ebook, comic, PDF, or audiobook.

Correct-by-construction TypeScript standards. Use for TypeScript engineering or when another skill needs the user's coding standards.

Design Effect services. Use when designing a new Effect service module or auditing an existing codebase for service, Layer, and composition improvements.

Access and manage my FreshRSS instance. Use for RSS reading, search, recommendations, subscriptions, auto-categorization, uncategorized-feed cleanup, or subscribing to an email newsletter through Kill the Newsletter.

Run and interpret Grove workspace scale performance benchmarks using `benchmark-scale`, including baseline capture, regression comparison, and optimization loops. Use when user asks about performance, scale, regressions, startup/discovery speed, polling/update throughput, p50/p95 timings, or benchmark harness usage. Trigger phrases: "benchmark-scale", "workspace scale", "perf regression", "p95", "performance baseline", "issue 38".

Instrumentation best practices for Grove's NDJSON event logging. Use when adding/changing log events, event fields, or logging schema in runtime code. Not for replay triage, use `replay-debug` for trace analysis and divergence debugging. Trigger phrases: "how to log", "add logging", "add event", "instrumentation", "telemetry", "event schema".

Replay and analyze Grove debug-record traces. Use when debugging runtime regressions, polling races, interactive input bugs, state transitions, frame/render mismatches, or when user asks to analyze `.grove/debug-record`. Not for instrumentation design, use `logging-best-practices` when adding or changing log events/fields. Trigger phrases: "replay", "debug record", "analyze trace", "runtime race", "timing bug", "state mismatch".

Look up FrankenTUI framework APIs, patterns, and examples from the reference codebase at .reference/frankentui/. Use when asking about FrankenTUI APIs, widgets, layout, subscriptions, rendering, styling, hit testing, or any ftui-* crate. Trigger phrases: "frankentui", "ftui", "how does frankentui", "ftui widget", "ftui layout", "ftui subscription", "ftui runtime".

Use dev-browser for browser automation on NixOS. Invoke when user asks to test UI, automate browser interactions, take screenshots, or verify web app behavior.