Skip to content

What works today

Every other page describes how WilsonOS is designed to work. This page says how much of it you can use today. It is the single source for availability: other pages carry a one-line notice pointing here instead of their own status claims, so no page can read as more available than this table. Compiled from the task ledger, which stays authoritative; rows were last checked on the date in the footer.

Availability

Label Meaning
Available You can use it now, as described.
Limited Usable, with a stated limitation you should know about.
Awaiting verification Built and possibly deployed, but the acceptance evidence is not yet in hand. Not to be relied on.
Planned Designed or intended; not built. Nothing on this site commissions it.

A claim moves up only on evidence. A written design, a merged change or a reported deployment is never counted as verification; a dated test, probe or live run against the running system is.

Evidence type says what kind of check stands behind each row, because different checks prove different things:

Evidence type Establishes
Instruction audit Tooling read every flow, skill, or registry entry and found the rule present. It does not prove the rule fired at run time.
Runtime acceptance A dated test, probe, or live run against the running system met the criteria.
Runtime check An automated check compared a claim with the running system (a doc-claims check, a content check after a cutover). Narrower than an acceptance suite.
Live use, no dated record In daily use, but no dated acceptance record is cited. Treated as Available on that basis, flagged so the gap is visible.
Reported deployment Merged and reported deployed; no acceptance evidence yet.
Written design A design the owner has agreed. Nothing more.
What you can do Availability Limitation Needs Evidence type Evidence
Talk to a brand agent in a private conversation; see its flow → skill → tool progress; get question forms for decisions Available — An account with brand access Live use, no dated record Deployed 2026-09-14 (chat routing P0/P1) and in daily use
Shared conversations with explicit membership Awaiting verification Deployed in code; acceptance held by the owner until the privacy and authenticated-turn probes pass under the isolation setup — Reported deployment Implemented; deployed; not verified (396 P2/P3, held 2026-09-15)
Human-only rooms (no agent) Planned Proposed extension to the conversation model — Written design Intended
Your conversation resumes where it left off; work in flight is not lost Awaiting verification Per-conversation runtimes are live; per-conversation operating-system isolation and its recovery guarantees are the held item above — Reported deployment Deployed; not verified
Your agent knows you, your brand, and its plate at session start Available — A completed profile and brand essence Runtime check The boot bundle is checked against the running system (doc-claims check)
Private memory scoped to you and the agent; conversation memory never merged across conversations Awaiting verification Writers and scope checks are deployed; acceptance is part of the held 396 P2 work — Reported deployment Deployed; not verified
Skills and workflows catalogue; agent announces which it is using; cost gate before any paid step Available The gate is proven present in every flow, not proven to have fired on a dated run — Instruction audit Audit tooling on the flow standard finds the cost gate in every sampled flow
A scheduled run does paid generation without asking Not a capability Every paid run — scheduled or not — stops for explicit approval in its destination conversation. Capped standing authorisation is a separate product decision, not documented here — Instruction audit Intended rule, present in flows
Scheduled automations with a run ledger, scoped credentials, required destinations, and a watchdog Available Scheduled runs use a process-environment boundary; the per-conversation operating-system boundary chat has is not yet applied to them Entries need a declared brand scope and a destination Instruction audit + runtime check Registry sync refuses unscoped entries (configuration audit); the watchdog alerts on missed fires (runtime)
Image / video / audio generation through one door, with a receipt opened at submit Available — Generation provider connections Runtime acceptance The media live suite (14/14, 2026-09-14) exercises the door and receipts
Publish gate — a task cannot be marked review/done while a media file has no completed publish Available warn exists only as an explicit, logged opt-out with a stated reason — Runtime acceptance Demo flow run on production 2026-09-12 (recorded report); default flipped to enforce 2026-09-14 on the owner’s direction; live suite 14/14 under the enforcing default; final dry-run 0 of 321 tasks refused (docs/reports/2026-09-14-media-phase4-flip.md)
Artifacts identified by stable id; rename without breaking links; durable primary storage with backup Available Some older path-keyed readers are still being moved to identity-based references — Runtime check Cutover reported and content-checked (397 Phase 6e, 2026-09-15)
Task folders keep identity when retitled Limited Reported live; final deployment and link checks pending — Reported deployment Deployed; verification pending (385)
Workbench: open any artifact type; comment on a version; rate; approve / redirect / reject Available — — Live use, no dated record Deployed and in daily use
Ratings attach to the specific artifact version, linked to the run and skill version Limited Today ratings are recorded at task grain (0 of 147 reviews carry a skill or artifact). The intended grain — the artifact version for media and non-media alike, rolling up to task summaries — is written, not implemented; non-media artifacts have no artifact-grain field yet (task 321 open) — Written design Intended; partially implemented (columns exist, writers thin)
A brand’s agent cannot act on another brand through a provider Limited Brand is logged on every gateway call, not yet refused — Reported deployment Deployed logging; enforcement intended
Connections catalogued with scopes; recorded external identifiers on publish Available — Per-service credentials Runtime acceptance The publish path is covered by the 2026-09-14 live suite
Publish a page from selected outputs (a collection, a client delivery page, a portfolio page) Limited An agent-assisted service: the agent composes the page from selected artifacts, you review it, the agent deploys it and records the URL. There is no built-in self-service publishing interface. A published page is a fixed snapshot of the versions chosen at publish time An agent conversation; a hosting connection Live use, no dated record The deploy skill is used for the system’s own pages; no dated customer-page record
Private client review pages — a recipient who can view, comment on, approve, or download a specific version, with access you can revoke Planned Hosting a page gives none of this: no recipient identity, no per-version feedback, no revocation. External feedback today arrives through you (chat, email) and you record it on the task — Written design Intended (share model on Share, export & publish)
Lead capture on a published page (a form, a download in exchange for an address) Planned No form processing; and attribution of a lead to a page needs the tracked-link registry that is not yet in place — Written design Intended (link registry — task 165)
Business analytics: advertising cockpit on the three-layer model Available First channel engine only; others at earlier maturity Advertising and royalty connections Live use, no dated record Deployed and in daily use; nightly factual refresh on the run ledger
Business finances — revenue, advertising spend, and a contribution figure per brand Limited Royalties and advertising spend are read from their providers and reconciled into the advertising cockpit. A named revenue store, goals computed from actuals, and a brand finances view are not built; other costs are not connected, so no figure on the system is a profit Royalty and advertising connections Reported deployment Reads and reconciliation deployed (task 164 re-scope, 2026-08-29); store and wiring intended
Cross-channel funnel (discovery → arrival → lead → conversion opportunity → sale) Limited One fixed five-stage funnel per brand is the measured model; the join across systems needs a tracked destination link that is not yet in place, so most paths report “not yet measurable” Tracked links per brand Written design Intended contract; partially instrumented
Per-business funnel designer (stages that vary: bookings, checkout, consultations) Planned — — Written design Intended (capability ladder)
Personal area: priorities, goals, commitments, everyday responsibilities, coordinated by Mia Limited Mia’s goal cadences (morning priorities, weekly interview, weekly review) exist; commitments and everyday-responsibility tracking are intended. The ownership model — same records with a personal scope, or a separate store — is an open decision — Live use, no dated record Cadences deployed on the run ledger; the rest intended
Recurring work and trackers (cadence → occurrences → per-occurrence tasks → periodic review) Planned Scheduled cadences and tasks exist separately; the tracker view that joins them is intended — Written design Intended
Mini apps (focused controls over one skill run) Limited Two guided surfaces are built (a thumbnail generator, a character-sheet generator); the general mini-app catalogue is in progress — Reported deployment Deployed (241/253 slices A–C)
Studio projects (multi-step production on a task’s scene tree) Limited Core shipped — scene tree, fan-out, assembled preview, elements, gallery folded on tags; the batch review grid and the agent-opened rich surface are not built; a visual polish pass is outstanding — Reported deployment Deployed (241 P0–P2, P4/P6/P7 by 253/256/270/272; reconciled 2026-08-15)
Nightly consolidation, rating recompute, drift detection; weekly review with owner-approved promotion Available Skill Showdown — the current skill against a candidate on the same brief, scored, then promoted by a person — is built (2026-09-21); batching several proposed upgrades into one showdown is not confirmed — Runtime check Scheduled runs recorded on the ledger
Access granted on the brand; row-level enforcement on every content table Limited Single operator today; owner columns and database-enforced access are being extended table by table — Reported deployment Partially implemented
People and teams — an organisation with members, roles, invitations, and clients distinguishable from teammates Planned Users, brand-scoped access grants, sign-in, and conversation membership checks exist for the single operator. The organisation record, invitations, and roles beyond brand access are designed, not built. The intended commercial model is per user; pricing is undecided; whether every external client reviewer must be a paid member is an open decision — Written design Intended (People and teams)
Multiple customers (tenants), each with isolated agents and billing Planned Designed; not scheduled — Written design Intended
Outside events wake an agent — a customer reply, a spend alert, or a form fill recorded and routed to the right agent or engine Planned No inbound event door; only you or a schedule starts work today — Written design Intended (System map, target)
Engine cockpit template — turning on a new engine gives you its cockpit Planned The two built cockpits share most of their tabs, but each was built separately — Written design Intended (engine contract)
Brand-wide shared memory, every entry stamped with who added it Planned Memory is scoped to one person and one agent today — Written design Intended (System map, target)
Trust ladder for skills and flows — propose only, do it and tell me, just handle it Planned How much a skill may do alone is set by hand in each skill; there is no battle-tested threshold that moves it up — Written design Intended
  • It will not inherit a status from a conceptual page. If a page reads as more available than this table says, the page is wrong and gets fixed.
  • It will not count a document, a merged change, or a reported deployment as verification.
  • It will not fold new proposals into in-flight completion work. Finishing the foundation is tracked separately from the next steps.

Sources: docs/tasks/INDEX.md and the open task docs (396, 397, 405, 409, 321), docs/reports/2026-09-14-media-phase4-flip.md, docs/architecture/agents-runtime.md §Residuals, docs/architecture/skills-and-flows.md §Skill Attribution, docs/architecture/measure.md §4–5; docs/wilsonos-app/capabilities.md (Funnel designer), docs/tasks/164-publisher-champ-kdp-ingest.md (re-scope 2026-08-29); docs/tasks/241-studio-ui-e2.md (reconciliation 2026-08-15), docs/architecture/identity-access.md §3, §7, input/WilsonOS_Overview.html §08 · Last verified 2026-10-06