Use when the user says `/orchestrator-mode`, asks you to coordinate agents, or gives large multi-phase work needing parallel delegation, reviews, PRs, scratchpads, or handoffs. Inputs - project goal, repo context, Solo agents/processes, branch/worktree constraints, locked user decisions. Do not use for a direct small edit, a normal code review, or single-agent implementation; use focused coding/review skills instead. Produces delegation plan, agent briefs, monitoring protocol, review/merge routing, state hygiene rules. Escalate on unclear scope, product direction, merge authority, destructive cleanup, or tool availability. Delegates are Solo processes: prefer Solo MCP tools, falling back to the `solo` CLI only when MCP is unavailable - never the in-process Task/Agent tool.
Use when initializing Rust best-practices rules in a project, the user invokes `/init-rust`, or a new Rust workspace needs convention rules wired up for an agent. Inputs - target project root, an existing or creatable `.claude/rules/` directory, and confirmation the project targets Rust edition 2024 (rules assume 2024 + MSRV 1.85). Do not use when the user wants language-agnostic project init, framework-specific scaffolding (axum, tauri, leptos, etc.), or actual code generation; use those skills or write code directly. Produces `.claude/rules/rust.md` with `paths: "**/*.rs"` frontmatter covering workspace layout, error handling, async/Tokio, CLI (clap), serde, tracing, type system, process supervision, security, testing, naming, tooling, and Edition 2024 changes. Escalate if `.claude/rules/rust.md` already exists with manual edits, the project is not a Rust workspace, or the user targets an edition older than 2024.
Use when the user asks to list Apple Calendar calendars or events, create/update/delete events, search appointments, or manage recurrence in Calendar.app on macOS. Inputs - calendar name when needed, event UID for read/update/delete, event title, start/end times, location/description/recurrence details for writes. Do not use when the user needs Google/Outlook calendar APIs, cross-platform sync, or general scheduling advice; use a service-specific calendar integration instead. Produces Calendar.app script commands and parsed event/calendar results. Escalate if calendar permissions are missing, the target calendar is read-only, dates are ambiguous, or deleting a recurring event could remove an entire series.
Use when the user asks to navigate a website, click UI, fill forms, take screenshots, scrape visible page data, test a web app, log into a browser session, or automate browser workflows with persistent page state. Inputs - target URL or existing page context, desired browser action, selectors or visible labels when known, credentials only when user explicitly provides them. Do not use when a plain HTTP request, local file edit, or Playwright test suite is the better artifact; use those tools instead. Produces dev-browser CLI commands or sandboxed JavaScript automation steps plus captured results. Escalate if login, payment, destructive actions, CAPTCHA, or sensitive personal data is involved.
Use when the user says "/meeting [title]", "/meeting end", or wants live meeting note capture from raw fragments during an active meeting. Inputs - meeting title, type, attendees, active meeting state, and real-time notes or end command. Do not use when the user wants post-hoc meeting summarization, a transcript cleanup, or general note-taking outside meeting mode; use summarization or writing skills instead. Produces structured raw notes, action items, decisions, and a shareable meeting summary for review. Escalate if no title is provided, no meeting is active on end, another meeting is already active, or content appears sensitive enough to require confirmation before writing.
Use when the user says "/record [message]" or when a meaningful workday event needs a concise timestamped entry in today's running log. Inputs - exact message for manual entries, current date/time, daily note location, and enough context to judge whether an automatic entry is meaningful. Do not use when the user wants a full journal entry, meeting notes, or end-of-day synthesis; use writing, meeting, or evaluate-day instead. Produces one deduplicated timestamped Running Log entry in today's daily note. Escalate if the daily note path cannot be created, the entry would duplicate existing content, or automatic logging would expose sensitive content.
Author or revise an AI-agent skill. Use when the user says "let's make a skill for X", invokes /skill-creator, asks to revise an existing SKILL.md, or wants to skill-ify a repeated workflow with 2+ prior invocations. Inputs - trigger phrase, artifact, evidence tier (E empirical / P practitioner / H heuristic). Do not use when the workflow is one-shot (write a snippet), is a phrasing trick or persona without a stable model, is an eval/benchmark loop (separate kit), or duplicates an existing skill (merge). Produces a skill directory with SKILL.md (5-part trigger description, lean body, no README/INSTALL/CHANGELOG) plus the registry-integration step for the active backend (see references/registry-integration.md). Escalate if no repeated workflow exists (suggest snippet), the model is H-tier with overreach ≥4 (manual-only), or the skill would compete with a stronger existing one (merge).
Use when the user asks to clean up a project, remove old plans, tidy docs, cleanup stale files, prune completed specs, or inspect artifacts after shipping a major feature. Inputs - project root, current branch context, and user confirmation before deleting or rewriting files. Do not use when the user wants disk-space cleanup, dependency pruning, code refactoring, or documentation generation; use a system-cleanup, package, refactor, or docs workflow instead. Produces a cleanup report classifying artifacts as remove, update, keep, or review, then confirmed file/doc changes. Escalate if deletion risk is ambiguous, living docs are involved, or more than two docs need substantive updates.