Roadmap

Roadmap

OmniRank version 0.4.0 ships two working Claude Code skills, audit and geo-artifacts, plus the omnirank fix CLI diff preview. Six further skills are designed and specified but not yet built. This page separates what exists today from what is planned, so nobody adopts the project expecting capability that has not been written yet.

The ordering below reflects one maintainer's judgement about sequence, not a commitment to dates. It changes in response to what people actually ask for in Discussions, and has already changed once — see "Why the targets below have already slipped" below.

Shipped in v0.4.0

Skill / command	What it does
audit	Scores a site's SEO, AEO, GEO, perf and security gates against its real HTML — including a site-level cross-URL pass and an indexability-contradictions pass — and emits a prioritised fix list with observed / expected / fix / fixTier per finding. Runs with no configuration beyond a URL. See Audit-Skill, Security-Layer, Contradictions.
geo-artifacts	Generates llms.txt, llms-full.txt and facts.json, each carrying an explicit citation licence (no licence granted by default as of v0.2.1). Python crawl path works on any stack; Node path imports a site's own content layer. See GEO-Artifacts-Skill.
omnirank fix (CLI, not a registered skill)	Resolves a finding's URL to its source file and prints the diff for the 4 mechanical finding ids, when locator confidence and blast radius both allow it. Writes nothing — no --write flag exists. See Fix-Preview.

Supporting surface that also ships: the omnirank.config.json schema (Configuration-Reference), the shared report format including fixTier and notEvaluated (Report-Schema), the finding registry and fix-tier/applicability model (Finding-Reference, Fix-Tiers-and-Applicability), the framework-aware locator (The-Locator), a CI-usable exit-code contract (CI-Recipes), and a packaged skill archive attached to every release.

What v0.4.0 added, specifically: a security layer (8 new finding ids: four inventory-only header gates, mixed content split by blockability, an http→https redirect check); five indexability-contradiction gates provable from a site's own declarations; Google rich-result required-property validation, distinct from schema.org validity; four on-page accessibility-overlap gates; and a corrected per-layer scoring model where a layer's budget scales with its own gate count instead of a flat constant. 67 finding ids total (up from 48), across 42 --fail-on gate names (up from 28). See the 0.4.0 changelog entry for the complete list.

Planned

Skill	Target	What it will do
aeo-onpage	v0.2 (slipped — see below)	Emit JSON-LD selected by entity type, draft and review AnswerBlocks, build FAQ blocks. On Next.js, emit typed builders and generateMetadata. Will require statistics, quotations and cited primary sources when drafting, per Research-and-Evidence.
indexing	v0.3	IndexNow submission, Google Search Console sitemap submit and URL Inspection, Bing Submit API, Wayback archiving, and hash-based freshness so dateModified only moves on real change. Will refuse to wire the Google Indexing API for general pages — that API covers JobPosting and BroadcastEvent only.
offsite-entity	v0.4	Read sameAs nulls as an entity-gap worklist, verify each claimed profile resolves and links back, run a peer mention-gap analysis, and draft outreach for manual send. Will never send email.
measure	v0.5	Rank tracking plus genuine AI-citation testing: query ChatGPT, Perplexity, Gemini and Copilot with a fixed question set and log whether the brand is named.
smm-content	v0.6	Turn a published URL into platform-native assets per configured locale, written to a review queue. Nothing enters the queue pre-approved.
smm-publish	v0.7	The only skill that writes to the public internet. Dry-run by default, requires human approval in the queue file, an explicit --confirm, and in-conversation confirmation before any live post.

Beyond v1.0: framework adapters for WordPress, Jekyll, Shopify, Astro, Vue and Svelte gain their own locators, alongside next-app-router, static, hugo and jekyll, which the locator already resolves as of v0.3.0 — see The-Locator. File-write capability (omnirank fix --write) ships once the locator is proven against real repositories and the write guarantees described in docs/research/2026-08-04-automation-architecture.md §2.5 are implemented and tested — not on a release number. v0.4.0 deliberately spent its budget on broadening what audit covers instead: auditing better is zero-risk, and a richer audit earns the right to edit files later.

Why the targets above have already slipped

Three targets have slipped so far, not one. aeo-onpage's "v0.2" target predates this table: 0.2.0 shipped as a site-level-gates / script-aware-AnswerBlock-bands / perf-layer release instead, folded into the existing audit skill rather than shipped as a new one. v0.3.0 then shipped the locator and omnirank fix — also not on this table when it was first written. offsite-entity's "v0.4" target has now slipped too: v0.4.0 shipped, and none of offsite-entity's sameAs gap analysis or peer mention-gap detection is built — the release spent its budget on the security layer, the contradiction gates, and the scoring fix instead. Treat every target version here as directional, not a commitment; none has a firm date, and the project has repeatedly shipped work that wasn't on this table ahead of work that was.

Why the order

aeo-onpage still comes first, though its job is narrower than it once was: omnirank fix already closes the loop mechanically for seo.canonical.missing, seo.canonical.relative, seo.canonical.chained and seo.schema.no-context — 4 of the 67 finding ids, all mechanical-tier. aeo-onpage's remaining job is the drafted-tier surfaces no mechanical generator can touch responsibly: titles, descriptions, H1 text, and AnswerBlock/FAQ content, all of which need judgement a human must approve per item. See Fix-Tiers-and-Applicability for why that's a hard boundary, not a temporary gap.

indexing precedes the off-site work because a page absent from an engine's index cannot be cited regardless of how well it is marked up — see the retrieval-overlap finding in Research-and-Evidence.

smm-publish is deliberately last and deliberately isolated. It is the only planned skill that can take an irreversible public action, and social platform APIs change without notice. Shipping it late means breakage degrades to a no-op rather than a wrong post.

Influencing it

This ordering is a guess until people disagree with it. The roadmap discussion asks three questions worth answering: which single skill would make you adopt this tomorrow, which one would you never use, and what is missing from the list entirely.

Contributions are welcome. Each skill is a SKILL.md, one or two reference documents, and Python behind it. The non-negotiables are in CONTRIBUTING.md — chiefly: never fabricate data, and never report a pass for a gate that did not run.

---

See also: Home · Audit-Skill · GEO-Artifacts-Skill · Fix-Tiers-and-Applicability · Research-and-Evidence