T0mSIlver

claude-remote-control-server

Install, update, inspect, or troubleshoot persistent Claude Code Remote Control servers for repositories. Use when the user asks to run `claude remote-control`, create a repo-specific remote-control server, make it survive reboots, manage user systemd services such as `claude-rc-skills.service`, configure `--spawn worktree`, service names, session names, prefixes, capacity, permission modes for spawned sessions (`--permission-mode`, `bypassPermissions`), lingering, or add another repo to claude.ai/code remote control.

delegate-to-claude-code

Call the Claude Code CLI (`claude`) as a delegated reviewer, second opinion, or edit worker. Use for Claude Code subagent-style delegation with `claude-rc-spawn`, `claude -p`, `--bg`, `--worktree`, `--tmux`, plan/read-only reviewers, acceptEdits/auto edit workers, JSON/stream-json capture, resume flows, and gotchas around permission modes, worktree cleanup, agent personas, and `--dangerously-skip-permissions`. For persistent `claude remote-control` repo servers, use the claude-remote-control-server skill instead.

delegate-to-codex

Call the OpenAI Codex CLI (`codex exec`) non-interactively to get a second opinion, run a code review, or delegate read-only or edit-capable work to an independent Codex run. Use for Codex CLI subagent-style delegation, long-running worker runs, worktree-isolated edits, machine-readable JSONL output, resume flows, and harness gotchas around stdin and sandboxing.

delegate-to-opencode

Call the opencode CLI (`opencode run`) non-interactively to get a second opinion, run a code review, or delegate read-only or edit-capable work to an independent opencode run. Use for GLM-5.2/opencode harness delegation, primary/all agent configs with edit permissions, worktree-isolated edit workers, JSON event capture, session resume/fork flows, and gotchas around `mode: subagent`, `--auto`, prompt files, and `--dir`.

drive-codex-in-herdr

Drive an interactive Codex CLI TUI in a background herdr pane — multi-turn delegation where you keep prompting the same live session, answer its approval prompts, and watch it in a pane. Use for herdr-based codex delegation, `herdr agent start/prompt/wait/read`, conversational follow-ups on one codex thread, approval gates that need answering mid-run, and long-lived workers that outlive this session. For one-shot non-interactive runs, use delegate-to-codex instead.

fastcontext

fastcontext is your default code-exploration subagent — it greps, globs, and reads a repository for you and returns file:line citations, keeping that exploration out of your own context. Invoke it via bash before answering, editing, reviewing, or debugging code you are not already certain about, and whenever the answer needs more than one file or tracing logic across modules. When in doubt, run fastcontext first.

unslop

Cut AI tells from any writing. Must always apply.

domain-modeling

Build and sharpen a project's domain model. Use when discussing codebase terminology, writing or editing a CONTEXT.md, or recording or editing an ADR.

Code & Development Listed

unnarrate

Remove self-narrating meta-commentary — text that describes the artifact it lives in, restates what the reader already sees, or narrates the process that produced it — from UI copy, docs, code comments, commit messages, and PR descriptions. Apply whenever writing or reviewing those surfaces, and before unslop when editing prose. Triggers: tooltips restating labels, "This section describes...", comments restating the next line, "progress is shown while it runs", PR bullets restating the diff.

grilling

Grill the user relentlessly about a plan, decision, or idea. Use when the user wants to stress-test their thinking, or uses any 'grill' trigger phrases.