Explains how switchroom works internally — config cascade, profiles, settings resolution, agent lifecycle, plugin system. Use when the user asks 'how does switchroom work internally', 'how does the cascade decide', 'which settings apply', architecture, design, or internals. Do NOT use for onboarding or getting-started questions ('how do I get started', 'I'm new to switchroom', 'bootstrap from scratch', 'set up for the first time') — those belong to switchroom-install.
Use when starting substantive development work — a code change, refactor, bug fix, infra change, or multi-step debugging task in any repo. Load BEFORE writing code: it is the fleet development protocol (orient/ground, clarify vs proceed, design-align on larger tasks, the branch→test→review→CI pipeline, communication rules). Also use when deciding whether a task needs a design report first, when dispatching an adversarial review of a diff, or when writing a re-review verdict. Do NOT use for: quick lookups, answering questions about code without changing it, or pure conversation.
Use ONLY when the user is asking the AGENT ITSELF about its own runtime state in a specific runtime-context — i.e. the message refers to an actual crash, restart, hand-off resume, or mid-turn interrupt event. Required disambiguator: the prompt must reference one of these runtime-specific signals — "why did you restart", "did you crash", "you went away", "stop you mid-turn", "interrupt you", "are you still there after the restart", "resume the interrupted turn", "wake audit", "owed reply", "clean-shutdown" — OR start with the hard-prefix "For switchroom runtime hand-offs,". Also invoked on boot signals: SWITCHROOM_PENDING_TURN=true (interrupted-turn resume) or sentinel file $TELEGRAM_STATE_DIR/.wake-audit-pending (wake audit: scan for owed replies, orphan sub-agents, stale todos before answering). Triggers on phrasings like "Why did you restart, please.", "you went away.", "can I stop you mid-turn.", "why did you restart.", "how do I interrupt you", "did you crash?", indirect signals like "the switchroom-runti
Cut and ship a switchroom release end-to-end: CHANGELOG consolidation, tag, npm publish, image build, and the fleet rollout gate. Use when the user says 'cut a release', 'ship a release', 'release vX', 'publish a new version', 'roll out the latest', or otherwise wants the merged work on main to go live for the fleet. This is the ONLY skill that authorizes a fleet rollout, and it enforces the npm-publish + image-build gates that were historically skipped (v0.18.4/v0.18.5 shipped to the fleet but never hit npm). Do NOT use for adding agents (switchroom-manage), diagnostics (switchroom-health), or a plain `switchroom update` on one agent (switchroom-cli).
Review the agent's OWN Hindsight memory bank with the most capable model and PROPOSE well-formed mental models ("knowledge models") to the operator via the approve/deny proposal card, instead of blindly creating them. Use to curate, seed, or audit the standing models the agent maintains over its specialty. Triggers on phrasings like: "curate mental models", "propose knowledge models", "review my memory bank", "what standing models should I have", "bootstrap mental models", "suggest mental models from my bank", "audit my mental models", "what recurring questions do I keep re-deriving", and typos like "propose knowlege models". Do NOT use for storing a single fact/preference/decision — that is retain, not a mental model. Do NOT use for identity or "who is the user" — that lives in profile banks, so never propose an identity model here (the operator's review of the card, not any code guard, is the only gate — so it's on you to honor). Even when the operator directly asks you to CREATE one specific named model wh
Run switchroom CLI operations on existing agents: logs, update, restart, version, config inspection, scheduled tasks, and Telegram plugin reference. HARD PREFIX TRIGGER: whenever the user's message starts with the phrase 'In switchroom (the CLI),' — regardless of what follows — use this skill. That prefix is load-bearing and wins over `switchroom-health`, `switchroom-runtime`, and `switchroom-manage`; even probes like 'In switchroom (the CLI), Can you why did it crash?', 'In switchroom (the CLI), Please sync my config.', and 'In switchroom (the CLI), Upgrade switchroom, please.' MUST route here, not to switchroom-runtime or switchroom-health. Use when the user wants to: show logs ('logs', 'what happened', 'check the journal', 'why did it crash'); update or upgrade agents ('update', 'pull latest', 'get new code', 'upgrade switchroom', 'Upgrade switchroom, please.'); restart agents ('restart', 'reboot', 'bounce', 'kick', "it's stuck"); check what's running ('version', 'what version is running', 'what sha', 'are
This skill should be used when the user asks to "test telegram", "test bot interactions", "mock the bot api", "write a telegram test", "test what users see in chat", "test progress card", "test the slot banner", "test soft-confirm", "test auto-fallback notification", "test pin behavior", or any variation on validating switchroom's Telegram-side output deterministically. Also use when the user mentions fake-bot-api, update-factory, bot-api.harness, GrammyError, e2e telegram, telegram regression test, or asks how to add a test for code that calls bot.api.* or handles incoming Telegram updates. Do NOT use for live progress-card behavior or telegram-plugin runtime questions — this is strictly for writing Bun tests under `telegram-plugin/tests/`.
List running switchroom agents with their uptime, model, and per-agent state. Strictly the "what's running and for how long" snapshot — nothing about install, restart, health, version, or update. Triggers ONLY on natural phrasings about listing/snapshotting agents and uptime, including: "Can you show me the fleet?", "Show me the fleet, please.", "Let's list switchroom agents.", "Let's how long has X been up.", "I need to how long has X been up.", "I'd like to what's the uptime of each agent.", "any way to list switchroom agents?", "quick q — can i show me the fleet", "pls per-agent snapshot", and typo'd variants like "per-agent snapshot", "per-agents napshot", "list swtchroom agents". Also fires on indirect signals like "how's the fleet doing", "what's alive right now", "is anything running right now". Do NOT use when the user is asking about anything OTHER than a running- agent list / uptime snapshot. In particular: - "fresh install / bootstrap / first-time setup" → `switchroom-install`. - "start / stop / r