Roadmap status
Updated 2026-09-11. This page distinguishes supported reading from repository workflows still under acceptance and future plans. Milestones are sequencing, not delivery-date promises; they do not authorize installation or account actions.
Canonical source: docs/ROADMAP.md
Work sequence#
| Stage | Focus | Tracker |
|---|---|---|
| Completed foundations | Readers, CLI, docs, hardening and scoped design/spike outcomes | M0 |
| Current verification | Implemented source skills, independent harness acceptance, browser QA and release checks | M1, #42 / #54 / #57 / #91 |
| Next | Inbox scheduling evidence, LinkedIn independent evaluation, bridge pairing/permissions | M2, #55 / #56 / #58 |
| Later / exploration | Visual jobs, capture benchmarks, more saved sources and separate companions | M3, #44 / #93–#96 |
Closing a design or spike issue only completes that scope. It never ships a production integration.
Current-main verification receipt#
On 9296ac5, an isolated Windows acceptance pass copied the primary and compatibility skills into temporary roots, generated the deterministic local fixture, and completed inspect -> preview -> read with captions and three frames. Preview required neither cloud approval nor a model download; read produced a manifest, three timestamped frames, and a non-empty sidecar-grounded transcript with no warnings. The merge commit also passed Linux 3.10, Linux 3.12, Windows 3.12, and Pages deployment.
This confirms the shipped reader's current installation and deterministic local path. It does not certify selected real inputs, private-source workflows, independent harness behavior, bridge security, provider adoption, or a final release/tag.
Capability evidence#
| Capability | Status | Evidence |
|---|---|---|
Guided inspect -> preview -> read | shipped | skill/scripts/voidscape.py; CLI tests |
| Local/URL video and audio evidence | shipped | skill/scripts/video.py; video tests |
| Local image and carousel evidence | shipped | skill/scripts/image.py; image tests |
| Local articles, RSS/Atom, approved public article fetch | shipped | skill/scripts/article.py; article tests |
| Optional screenshot provenance | shipped | Image sidecar contract, merged PR #107; local hash/geometry verification, not browser capture or producer authentication |
| Public feed capture and notes | dev-only | RSS intake helper, selected article/enclosure reads (#109–#111), real podcast QA and one public access-wall skip; project skill and verified publication implemented; broader provider and independent harness acceptance pending |
| Local chat export (WhatsApp-style) evidence | shipped | skill/scripts/chat.py; chat reader tests |
| Connectors contract page | shipped | this tree: connectors.md |
| Harness skill kit (copy-and-adapt templates) | shipped | harness/skills/; template tests |
| Shared reader protocol and envelope | shipped | reader manifests; protocol tests |
| Private recovery pointers and Windows UTF-8 output | shipped | merged PR #65; recovery and Unicode subprocess tests |
| Bounded article responses and direct loopback health | shipped | merged PR #66; network boundary tests |
| Source capability registry and reader override | shipped | merged PR #45; supported voidscape route, voidscape sources; source tests |
| Instagram URL queue helper | dev-only | repository script; not installed with the skill |
| Instagram triage and verified notes | dev-only | merged project skill/controller and note store (#77–#80); independent behavior and live harness acceptance remain #54 |
| LinkedIn observed-post capture and notes | dev-only | local workflow: typed identities, verified notes/index, legacy assessment and project skill with selected resume; independent harness evaluation remains #55, while any live platform work needs a documented approved route |
| Instagram follow-relationship audit (read-only) | dev-only | repository script; local exports only; never unfollows |
| YouTube private-playlist queue adapter | dev-only | repository script; official API; not installed with the skill |
| Public YouTube capture and notes | dev-only | source workflow; real local capture/read/note proof, independent skill/harness acceptance pending |
| Agent documentation tree | shipped | this page and sibling pages |
| Machine-readable agent discovery manifest | shipped | docs/agents/manifest.json |
| Observe-and-capture playbook | shipped | docs/agents/observe-and-capture.md |
| Thin observe CLI | shipped | skill/scripts/observe.py; issue #24 |
| Browser bridge production integration | dev-only | isolated spike complete; a pairing and permission design exists in the separate agent-bridge repository, while independent review and isolated client acceptance remain #58 |
| MCP host spike / decision | shipped | no-go report; issue #26 |
| Production MCP host | parked | revisit gates in the #26 spike report |
| Browser extension spike | dev-only | spike report: isolated Chrome three-command proof; real harnesses and signed-in profiles unverified |
| Codex plugin bundle | dev-only | validates locally; SkillSpector CRITICAL/DO_NOT_INSTALL gate remains open |
| Production universal browser extension | parked | separate security/repository decision required |
| Local recording inbox controller | dev-only | repository helper, project skill, local note author and a manual Windows scheduler-plan helper exist; real recording/harness acceptance and observed scheduled execution remain #56 |
| Unattended orchestration | parked | general orchestration beyond the bounded local inbox work in issue #56 needs separate product/privacy design |
| Hosted SaaS | parked | product, billing, connector, and legal gates required |
Interpretation rules#
shipped requires merged implementation, tests on main and a supported entry point. release-candidate means an implemented and verified change awaiting merge/release. Repository source workflows and the browser spike remain dev-only while supported distribution or independent end-to-end acceptance is incomplete. Project-local packaging is not live harness proof. A design spec or fake spike never upgrades a capability to shipped.
Iris scan warnings are not evidence of malware. Richard assesses them as false positives; the remaining gate is pinned capability/permission review and integration acceptance. Iris is not installed by this roadmap; its existing STOP remains an adoption status.
For phase-level history and deferred platform ideas, read the canonical roadmap. The open issue list is scheduling evidence, not product evidence.
Incoming — what exploration could unlock#
The long-term direction is one idea: every piece of media you already keep becomes inspectable, citable evidence through the same gates. None of the directions below is shipped, installed, or promised; each is listed with what it would take to become real.
| Direction | What it would give you | Standing gates |
|---|---|---|
| X bookmarks, TikTok favorites, Reddit saves | one evidence library across more platforms you save to | per-platform permission review; signed-in capture is never implied by public reading |
| Newsletter collections | inbox reading with sender and issue order preserved | delivery-platform review; no mailbox credentials, ever |
| Browser bridge (live pages) | read what a page shows at read time, with the page state recorded | site-by-site approval; no credentials, cookies, or storage access |
| Screenshot CLI/MCP integrations (Iris-style) | turn a visible screen region into citable frames | capture-scope consent; same inspect -> preview -> read discipline |
| Agent plugins and multi-model reading | choose per-task models for transcription and vision | model-download and cloud approvals stay per-run, never global settings |
| Hosted edition | Voidscape without local setup | product, billing, connector, and legal gates; local-first stays the default |
Two boundaries hold across every row. First, Voidscape is a reader: follower audits, creator performance dashboards, and job-search tooling are separate companion projects, and permission to read never implies following, unfollowing, messaging, or publishing. Second, security review gates apply before any exploration becomes an installed capability — a working prototype is not a release.