# The Muse Field Guide > A collection of resources for people getting started with Meta Muse, from one person who runs a large share of his digital day to day through it: guides, templates, checklists, skills, tips, and the lessons learned the hard way. Everything here is in daily use in the operator's own setup. Nothing is aggregated from other people's content. The guide is written by Nic (niclydon.io), whose Muse agent, Nova, holds a chief-of-staff seat in his setup. Every resource below has a markdown version at the linked URL; the HTML pages carry the same text. Published figures on the site carry a source and a verified date, and the footer of every page carries the facts-verified stamp. One framing note: the resources are written for a Muse user with none of the author's other infrastructure. A few pages describe machinery from the author's wider personal platform as context, and say so where they do; that machinery is not something Muse provides out of the box. ## Start here - [Start here](https://www.niclydon.io/muse/deep-dives/start-here.md): What this site is, who it is for, and where Muse ends. - [Steal this setup: standing files](https://www.niclydon.io/muse/items/standing-files-starter-kit.md): Step one. Four small files that say who your agent is, who you are, how work gets done, and what it should remember. - [One side chat per topic](https://www.niclydon.io/muse/items/one-chat-per-topic.md): Step two. The chat list is the org chart; the main chat stays a desk, not an archive. - [The Verify lane: one claim at a time](https://www.niclydon.io/muse/items/the-verify-lane.md): Step three. A standing lane where uncertain claims get checked one at a time instead of hardening into facts. ## Guides - [One side chat per topic](https://www.niclydon.io/muse/items/one-chat-per-topic.md): The chat list is the org chart: each side chat owns one project, device, program, or chore. - [Standing files as configuration](https://www.niclydon.io/muse/items/standing-files-as-configuration.md): Five plain files configure the agent, kept separate, amended in the open, each rule citing the incident that bought it. - [Watchers that stay silent](https://www.niclydon.io/muse/items/silent-watchers.md): A temporary job with a named expected state: silent while it holds, one report with evidence when it changes, then it deletes itself. - [The Engine Room: one chat for noise](https://www.niclydon.io/muse/items/the-engine-room.md): All routine machine output delivers to one designated chat; the main chat receives only escalations. - [The Verify lane: one claim at a time](https://www.niclydon.io/muse/items/the-verify-lane.md): Uncertain claims are brought to the human one at a time, evidence first, until the data is clean. - [The approval ladder](https://www.niclydon.io/muse/items/the-approval-ladder.md): Commit, push, publish, book, and send are separate approvals; none implies the next. ## Templates & checklists - [Steal this setup: standing files](https://www.niclydon.io/muse/items/standing-files-starter-kit.md): Four files to write today; every correction becomes a same-day edit. - [Steal this setup: the Engine Room](https://www.niclydon.io/muse/items/engine-room-starter.md): One designated chat for machine output, a registry for every standing job, and a weekly drift digest. - [Steal this setup: a Verify lane](https://www.niclydon.io/muse/items/verify-lane-starter.md): A standing chat and a queue of uncertain claims, evidence first, one question per message. - [The question pre-flight checklist](https://www.niclydon.io/muse/items/question-preflight-checklist.md): Seven checks the agent runs before any question may reach the operator. - [The public-editions scrub checklist](https://www.niclydon.io/muse/items/public-editions-scrub-checklist.md): The repeatable pass run before any private repository gets a public edition. - [The Engine Room registry template](https://www.niclydon.io/muse/items/engine-room-registry-checklist.md): Every standing job registered with a class, a delivery home, and an end date. ## Skills I've built - [ship: the one-word closing ritual](https://www.niclydon.io/muse/items/ship.md): One word runs the whole close: narrative record, commit, merge, deploy, verify live, clean up. - [where-were-we: re-ground a stale chat](https://www.niclydon.io/muse/items/where-were-we.md): Reconstructs what a thread was for, checks what is still open, names the owner, gives the single next step. - [upgrades-scan: ranked fixes, item verdicts](https://www.niclydon.io/muse/items/upgrades-scan.md): A read-only sweep returns a ranked list of opportunities, disposed item by item. - [narrative-docs-update: the record, in-repo](https://www.niclydon.io/muse/items/narrative-docs-update.md): After a coherent arc of work, the story goes into the work's own repository. - [baton-pass: hand a chat to a successor](https://www.niclydon.io/muse/items/baton-pass.md): Hands a side chat's work to a fresh successor chat, with a decisions file. - [sidechat-tidy: sweep and rename every chat](https://www.niclydon.io/muse/items/sidechat-tidy.md): Renames every chat under one naming system with status markers. - [verify-loop: audit memory one claim at a time](https://www.niclydon.io/muse/items/verify-loop.md): Works a queue of uncertain memory claims, exhausting every source family first. - [receipt-matcher: reconcile receipts to transactions](https://www.niclydon.io/muse/items/receipt-matcher.md): Matches bank and card transactions against email receipts; read-only by design. ## Tips & tricks - [Run a pre-flight before asking](https://www.niclydon.io/muse/items/preflight-before-asking.md): The agent checks every source it has before a question reaches the human; most questions die there. - [One question, at the end](https://www.niclydon.io/muse/items/one-question-at-a-time.md): One question per message, placed last, offered as tap options. - [Ask only for ground truth](https://www.niclydon.io/muse/items/ask-only-ground-truth.md): The human is asked only for facts no source can resolve, or a genuine preference or consent choice. - [Verify uptake by claims, not status](https://www.niclydon.io/muse/items/verify-uptake-by-claims.md): Work is real when a work product or recorded claim exists under the worker's name. - [Done means verified live](https://www.niclydon.io/muse/items/done-means-verified-live.md): Done is a live check in production, not a commit, a merge, or a deploy. - [Spot-check every background number](https://www.niclydon.io/muse/items/spot-check-background-numbers.md): Any aggregate an agent reports is a lead, not a fact, until a foreground check reproduces it. - [Probe capacity; never remember it](https://www.niclydon.io/muse/items/probe-capacity.md): Whether a lane or service is available is checked now, not recalled from yesterday. - [Never accept a fabricated absence](https://www.niclydon.io/muse/items/never-accept-fabricated-absence.md): "Not found" must state the scope of the search. - [Pre-register kill criteria](https://www.niclydon.io/muse/items/pre-register-kill-criteria.md): Write down what result would kill an experiment before it runs; honor it when it arrives. - [Ingest before you delete](https://www.niclydon.io/muse/items/ingest-before-you-delete.md): Nothing is retired until its history is proven to exist somewhere durable. - [Restate the question with the choices](https://www.niclydon.io/muse/items/restate-the-question.md): Every re-shown choice carries the plain-text question it answers. - [Convert refused steps into tracked work](https://www.niclydon.io/muse/items/convert-refused-step-to-tracked-work.md): A declined manual step becomes tracked work with the exact spec attached. - [Corrections land in files same-day](https://www.niclydon.io/muse/items/corrections-land-same-day.md): A correction agreed today is written into the standing files today. ## Learned the hard way - [Ask one question at a time](https://www.niclydon.io/muse/items/verdict-one-question-at-a-time.md): Tested daily in the verification lane and inbox triage; batches stall, singles get answered. - [Don't let a model decide what to ask you](https://www.niclydon.io/muse/items/verdict-model-as-judge.md): Every model tested judged ask-or-not at coin-flip level; replaced by rules. - [Turn judgment calls into checks over facts](https://www.niclydon.io/muse/items/verdict-rules-over-facts.md): Where criteria can be stated as checks over facts, state them as checks. - [Watchers should stay silent, then delete themselves](https://www.niclydon.io/muse/items/verdict-silent-watchers.md): A go-live watcher stayed quiet while the expected state held, then closed its check and deleted itself. - [Retire an unreliable model; don't demote it to fallback](https://www.niclydon.io/muse/items/verdict-unreliable-model-fallback.md): A fallback is a primary waiting for a bad day. - [Count work only when a claim or deliverable exists](https://www.niclydon.io/muse/items/verdict-verify-uptake-by-claims.md): Caught a worker reporting "working" while its pane sat parked on an approval dialog. - [Judge a connector by how it behaves unattended](https://www.niclydon.io/muse/items/verdict-connectors-unattended.md): A connector that raised four approval cards in five minutes stalled background work dead. - [Publish numbers with a source and a verified date](https://www.niclydon.io/muse/items/verdict-facts-with-dates.md): On the verifier's first cycle, 13 of 16 published facts had drifted. ## Deep dives - [Start here](https://www.niclydon.io/muse/deep-dives/start-here.md): What this site is, who it is for, and where Muse ends. - [Chats as an operating model](https://www.niclydon.io/muse/deep-dives/chats.md): One side chat per topic; the main chat as a desk. - [Standing files: configure your agent like you mean it](https://www.niclydon.io/muse/deep-dives/standing-files.md): Configuration as files you own. - [Memory that holds](https://www.niclydon.io/muse/deep-dives/memory.md): Curated files with provenance, a verification lane, file-first working. - [Grow your own skills](https://www.niclydon.io/muse/deep-dives/skills.md): 54 shipped skills, 63 grown at home. - [Practices and tips: the resource core](https://www.niclydon.io/muse/deep-dives/practices.md): Thirteen practices, each with the receipt that bought it. - [Steal this](https://www.niclydon.io/muse/deep-dives/steal-this.md): Public repos, checklists, and templates, written to be copied. ## Using this guide - [How this guide is meant to be used](https://www.niclydon.io/muse/how-to-use.md): The four uses this guide supports: explaining it, guiding reading, auditing a setup, and drafting an adoption plan. - [Self-audit rubric](https://www.niclydon.io/muse/self-audit.md): Seven dimensions for evaluating a Muse setup, with check questions and what good looks like. - [Adoption guidance](https://www.niclydon.io/muse/adoption.md): Sequencing, adapting rather than copying, and what to skip if a setup lacks the receipts behind a resource. ## Optional - [The whole guide in one file](https://www.niclydon.io/muse/llms-full.txt): The complete text of every resource and deep dive, plus the rubric and adoption guidance, in a single document. Large; useful for agents with a big context window, unnecessary otherwise. - [items.json](https://www.niclydon.io/muse/items.json): The machine-readable manifest of all 41 resources.