OMK's provider-neutral coding-agent CLI also exposes a multi-agent control plane for explicitly configured workflows. The diagram describes that design, not what every prompt automatically runs.

The v0.98.3 SDK rejects incomplete first-party judge responses and exposes deterministic ties; it is not an automatic TUI judge.

1. Scope the goal, paths, resources, and acceptance predicates. A selected orchestration workflow may also supply a DAG; an ordinary prompt remains one agent/tool loop.
2. Route work to models, agent skills, MCP tools, and extensions without changing the evidence contract.
3. Verify declared checks in explicit evidence workflows. Required failing checks block those workflows; advisory judging cannot replace them.
4. Preserve receipts and replay state for bounded session recovery; continue durable goals from explicit reducer state.

The animation changes once every 1.5 seconds and contains no flashing. The four steps above are the complete text alternative.

OMK keeps routing separate from control and evidence. Codex, Claude Code, OpenCode Zen/Go, Kimi, GLM/ZAI, native xAI/Grok, NVIDIA NIM, and local providers can participate through omk-ai while the run contract stays stable.

Native xai keeps subscription OAuth and XAI_API_KEY billing separate. See provider setup, provider resilience, and Grok integration.

Published packages

npm install omk-agent-core
npm install omk-ai
npm install omk-protocol
omk install npm:omk-book-to-skill@0.98.3
npm install omk-tui

v0.97.0 shipped the OpenWiki policy and workflow, but no versioned corpus or integrity checker. The following integrity/output guards shipped in v0.98.0; the generated corpus remains optional and is not bundled:

• openwiki/ — absent. The previous untracked corpus was removed after the hardened gate proved it carried fabricated evidence: 8 frontmatter symbols that no declared source path defines (AgentLoop, getModel, DeepWall, loadExtensions, createExtensionRuntime, main), 45 references to @omk/* package names this repository does not publish, and a restatement of this README's Scope -> Route -> Verify -> Replay loop as a strict engine state machine, which is not what the source implements. CI regenerates the corpus; nothing is lost.
• scripts/check-openwiki.mjs — shipped integrity checker. An interrupted corpus now fails unless openwiki/.manual-review.json binds a review to the exact corpus digest, and every frontmatter symbol must bind to one of that page's own source_paths as a whole identifier.
• scripts/check-openwiki-output.mjs — output gate. The scheduled workflow may write, upload, and open a PR for openwiki/ and nothing else, so a model reading this repository cannot reach AGENTS.md, CLAUDE.md, or the workflow that runs it. The gate runs once before the artifact leaves the read-only generating job and again before the PR, because the publishing job holds write permissions the first one does not.
• .understand-anything/ — optional local structural graph used by Pi Lens; it is not published or injected into prompts by default. To reach a session, attach it through OMK's MCP client like any other server; there is no second, bespoke path for it.

Source and tests remain authoritative. Shipped guards do not turn a generated index into authority: treat corpus pages as local advisory data and recheck source.

Retrieval

A corpus no session can read is documentation of a plan, not a feature, so the pages are now candidates for prompt budgeting. Enable contextBudget.openwiki alongside contextBudget.enabled (settings) and each page becomes a low-priority evidence item ranked against the turn's query. Pages compete for leftover budget and can never displace instructions or skills; most turns carry titles and declared symbols alone, and a page's text arrives only when the query earns it.

Admission mirrors scripts/check-openwiki.mjs rather than restating it. A complete corpus at the current HEAD offers page text; one whose HEAD has moved offers titles only and is marked stale; an interrupted corpus is refused unless a review binds to its exact digest. The default is off, and with the setting off the prompt is byte-identical to one built without a corpus.

The design decisions behind OMK's context, routing, memory, and orchestration layers are grounded in published work rather than invented in isolation. Each row below was retrieved and read directly; claims are at abstract level, which is the evidence grade this table asserts and no more.

Entries include implemented mechanisms and design proposals; check the runtime status guide for availability. The wider survey, including approaches not adopted, is working material that is not published with the repository.

Historical correction: the immutable v0.97.0 notes below announced a versioned OpenWiki corpus, but that release still ignored /openwiki/ and did not contain the corpus or checker. See the current repository-understanding section above for the shipped guards and optional-corpus boundary.

Release v1.0.0

Added

• Bundled Neo skills: six public skills ship in resources/neo/skills and load when no user, project, or explicit skill supplies a name (OMK_BUNDLED_SKILLS=0 or --no-skills to disable), plus an omk neo subcommand and an ACP conversation-only mode (--mode acp).
• Subagent lane execution settlement: lane results model settled/failed/unsettled, a shared permit pool tracks unsettled children across dispatch instances, heavy-lane admission narrows the width gate, and permit weights validate explicitly.
• In-memory run usage ledger (run-usage-ledger, run-usage-operation): reserves capped budgets per attempt, records transports and cumulative usage idempotently, binds settlement to the attempt admitted at entry (caller input mutation cannot redirect it), and retains reservations for unknown usage — accounting only, not a hard financial limit.
• terminal-browser extension example: ports the Claude Code plugin to OMK, rendering a real browser in a kitty-graphics overlay via the bridge HTTP API; /browser command and open/close tools.
• Strict-evidence approval adapter in guardrails/ binding protocol-level evidence reports to host approval decisions.

Fixed

• Empty streamed completions (success-shaped stop with zero usable output) classify as dead streams: they no longer reset the retry budget, pin the UI in a retrying state, or rotate to a live route in the same model family (ox-alpha/union-alpha rotation). Anthropic-messages base URLs carrying a version suffix are normalized so /v1/messages never doubles.

Added (previous cycle work)

• --thinking ultra is now a first-class Astra selector level. GPT-6 Astra maps it to the documented max effort rather than clamping it away or sending an invented ultra wire value.
• The status rail's USAGE section now covers commandcode: it calls Command Code's /alpha whoami, credits, subscription, and usage-summary endpoints with the stored API key and renders the 5-hour, weekly, and monthly credit meters with reset times, or the plan name when rolling windows are absent.

Fixed

• Missing ESM named-export errors such as does not provide an export named 'MAX_FRAME_BYTES' classify as configuration, not as an orphan tool-call protocol fault. The Next action tells the operator to restart OMK; /new does not reload provider modules.
• The status rail's Devin USAGE meters now match the CLI /usage surface: credit-billed plans no longer render proto-default 0% remaining as exhausted 1D/7D windows, and GetUserStatus unary gzip bodies decode.

Release notes live in RELEASE_NOTES_v1.0.0.md.

Release v0.99.0

Breaking Changes

• Checkout-only TB tooling now emits selectionVersion: 2 and omk-tb21-audit-report-2. Selection totals can be null when estimates are unknown, and audits reject missing or invalid start/finish timestamps. The input manifest remains omk-tb21-manifest-1. Update report consumers and freeze new task selections before comparing runs; see selection compatibility and audit rules.

Added

• Verified command DAGs accept an explicitly approved writer.maxConcurrentTasks: 2 and release ready dependants without an unrelated-task barrier. Omission keeps serial execution and legacy contract digests. Task-bound execution IDs and draining preserve cancellation/recovery boundaries; final verification still uses one fixed candidate. See Verified Run.

• Added the Devin SWE-2 harness: a devin-harness domain loadout auto-applied when the devin provider is active (OMK_DEVIN_HARNESS=0 disables it, independently of OMK_GROK_HARNESS), per-turn <active_skills source="devin-harness"> grants selected from the live inventory, an optional ~/.omk/agent/devin.md operator overlay, and the canonical Devin SWE-2 harness guide. devin/swe-2 now carries a 1,000,000-token local context budget that selects the account catalog's 1M-context lane; a lane that declares a smaller window fails the request instead of shrinking the budget.

• The status rail's USAGE section now covers devin: it calls GetUserStatus with the stored CLI session token (OAuth or DEVIN_API_KEY) and renders the plan's daily/weekly quota meters with reset times, or the plan name and credit balances when the account reports no quota windows.

Changed

• Grok and Devin harness dispatch share provider-harness-dispatch.ts (loadout runtime injected, outside the sdk.ts import cycle) and harness-skills.ts; the grok-harness and devin-harness profiles live in domain-loadouts-provider-harness.ts. tryGrokHarnessDispatch() and selectGrokHarnessSkills() keep their public signatures and behavior.

Fixed

• A background task result delivered with triggerTurn while the session was sleeping in retry backoff started a second top-level run: sendCustomMessage only checked isStreaming, not isRetrying (unlike prompt()). When the retry woke up, agent.continue() threw Agent is already processing, and the runtime-failure handler closed the run journal that the competing run still owned, so that run later died with run journal received agent_end without run_started and the session was wedged until a model switch. Custom messages now queue during retry backoff exactly as during streaming, and a run that never owned the journal (rejected with Agent is already processing while another run is live) no longer finishes or clears it.

• Codex SSE response-header waits now respect retry.provider.timeoutMs, with the existing 10-second minimum, rather than aborting every large request at a fixed 10 seconds. This does not extend an outer run deadline.

• Resource completion descriptions remove decorative leading [OMX]/[OMO] labels without changing source metadata, invocation names, or enabled tools; marker-only descriptions display OMK resource.

Release notes live in RELEASE_NOTES_v0.99.0.md.

Release v0.98.5

Added

• Added /debug runtime inspection, explicit metadata-only local reports with /debug save, and structured failure cards with expandable details. UI entry observations do not infer build revisions or authorize retries.
• Added opt-in verified-run CLI/SDK paths with protected verification, immutable candidate recovery, input-checkpoint writer restart, and static DAG task retry. These paths retain approval, ownership and budget boundaries; they do not apply artifacts to the original workspace automatically.

Fixed

• Strengthened execution ownership and shared-budget boundaries. Pre-commit checks preserve the selected index, including partially staged files, instead of expanding the commit.
• Execution-ownership wrappers retain lazy, context-sensitive tool timeouts and stale-context rejection instead of fixing the timeout at registration.

Release notes live in RELEASE_NOTES_v0.98.5.md.