Execution follow-up: 2026-09-07 source repairs, tests and runtime limits updates this dated audit view. The master register owns current task dispositions; earlier observations remain historical.
Status: Current source comparison; deployed compatibility not certified Reviewed: 2026-09-07. Earlier 2026-05/07 snapshots remain below as historical evidence.
The current view follows the estate audit, VisionClaw audit, Agentbox audit and federation audit. Working-tree source, declared configuration and a running deployment are separate evidence states. No earlier completion percentage, fixture result or tool count becomes a current result merely by appearing in this file.
| Area | Source-supported implementation | Remaining compatibility evidence |
|---|---|---|
| Identity and replay | VisionClaw has a single-use process-local NIP-98 cache (src/utils/nip98.rs); native Solid has a ReplayStore seam; forum/edge and Agentbox have their own authentication boundaries. Agentbox canonical identity is did:nostr:<64-hex> with Multikey material offered alongside, not a replacement DID string. |
Replay behaviour must be measured per consumer, restart and replica. Shared key material is not one shared verifier. VisionClaw NIP-26 user delegation remains deferred; forum mesh delegation types/verification do not establish its outbound integration. |
| Mesh | Forum nostr-bbs-mesh supplies transport/core types and is a relay-worker dependency; mesh.rs provides configuration and fan-out planning plus inbound admission gates. Agentbox/VisionClaw share an explicit versioned federation kind map. |
Relay accept-path/outbound connector joins remain deferred. Do not call the estate either fully federated or wholly scaffold-only. Deployment mode and enabled peers require a current instance receipt. |
| Pod | VisionClaw embeds a library, Agentbox selects a supervised native server, and the forum uses a separate Worker tier. Consumer versions/features differ from the standalone source. | Storage policy, cache, WAC, OIDC and replay guarantees require exact consumer wiring and tests. Library capabilities do not imply parity across all tiers. |
| Governance | VisionClaw includes the broker domain kernel, ACSP integration and inbox/decide REST surfaces; the legacy BrokerActor/Neo4j transport was deliberately left behind. Agentbox task spawn has a journalled local fast path. Forum decision projection has stronger transactional receipts. | Signed, accepted, projected, received and applied are separate stages. Request/case/operation correlation, guarded transitions and external application still need end-to-end proof. A task-spawn journal is not universal tool or nightly-loop confinement. |
| Ontology bridge | The current ontology-bridge.js TOOLS array has eleven advertised entries. ontology_propose has a dispatch branch but is not in that advertised array. Governed writes and bounded retrieval exist; the configured retrieval backend can be VisionClaw or Loom. |
Count the advertised registry at the selected revision, not a historic twelve-tool description. Loom client generation attestation, asserted/inferred scope and its own search/SPARQL recall measurement remain open. |
| Deployment and evidence | Nix/compose/CI gates and release manifests exist. Agentbox source fixes may be staged pending image rebuild. Current audit reports name tested boundaries and limitations. | Bind source/lock/features, image, projected config and loaded process. A green structural gate or historical fixture result does not certify the deployed estate. |
The six conditions of arXiv 2609.12482 grade every surface where a human decides on an agent's behalf. Status vocabulary: absent (no code or receipt), partial (mechanism exists but the condition is not measurable from it), measured (a runtime quantity or receipt exists). A status may be raised only by a file:line citation to code or a receipt, never by a document. Citations are to the merged main heads of the three substrates — nostr-rust-forum 11b674c, agentbox b859b37d2, VisionClaw 8f9affe49 — where PRD-augmentation-conditions FR2–FR7 landed. C1–C3 are snapshot conditions; C4–C6 are longitudinal, so a first measured status on those rows is a baseline, not a pass (ADR-2010 consequence 3) — it is this cycle's only measurement, and a second cycle (review_trigger: 2026-12-14) is what actually grades whether learning, pathways and job purpose are holding up over time.
| Condition | nostr-rust-forum | agentbox | VisionClaw |
|---|---|---|---|
| C1 Durable net value | partial — per-reviewer time-to-decision is now measured (crates/nostr-bbs-auth-worker/src/governance_api.rs:1287-1288) but nothing forum-side differences it against a productivity claim |
partial — dream-ledger reviewer minutes are measured per reviewer (management-api/lib/dream-ledger.js:203-222) but no net-value calculation consumes them |
measured — HITL Precision is a real computation over decided cases, no longer a stub (src/services/kpi_compute.rs:213, wired at src/services/kpi_compute.rs:408-420); Whelk-gate rows are excluded from the human-only series (src/services/kpi_compute.rs:140-154) |
| C2 Meaningful human control | measured — the proposal fields/reasoning render below the controls and the typed rationale is what publishes, never a template (crates/nostr-bbs-forum-client/src/utils/governance_view.rs:57,77,445); the reviewer-role admission gate gives delegated authority to override (crates/nostr-bbs-relay-worker/src/relay_do/nip_handlers.rs:234-249) |
measured — the fail-closed deny path names the manual-continuation tool at the point of denial, the code comment cites C2 by id (management-api/lib/authority.js:395-404); tightening is a lattice join the operator's manifest always wins (management-api/lib/authority.js:122-140) |
measured — the proposal renders unclipped and the rationale gate is enforced server-side, not just in the client (client/src/features/control-center/governance/AcspCaseQueue.tsx:16-24; src/handlers/enrichment_proposals_handler.rs:406-423,503-518) |
| C3 Accountability and recovery | measured — the application-receipt endpoint advances broker_decisions through consumer-received/applied/not-applied (crates/nostr-bbs-auth-worker/src/governance_api.rs:1124); the ageing cron mints escalated-on-age receipts past max_pending_hours (crates/nostr-bbs-relay-worker/src/cron.rs:595, wired at crates/nostr-bbs-relay-worker/src/lib.rs:1003); the client renders every ladder/terminal/conflicting stage (crates/nostr-bbs-forum-client/src/stores/receipts.rs:56-86) |
measured — every gate denial is appended to the hash-chained authority journal with stage and reason (management-api/lib/authority-journal.js:42,92-115); receipt-post failures are journalled, never silent (management-api/lib/governance-receipt-publisher.js:193-206,312) |
measured — ElevationActor gains a 14-day OPEN_CASE_TTL, a Reconcile message and an expired receipt path (src/actors/elevation_actor.rs:68,93,1099-1132) |
| C4 Deepening learning (baseline) | measured (baseline) — GET /api/governance/reviewers computes median/p90 time-to-decision, override rate and calibration/probe catch rate per reviewer; the code names C4 and C6 as the questions it answers (crates/nostr-bbs-auth-worker/src/governance_api.rs:1336-1394,1397) |
measured (baseline) — dream-ledger reviewerStats tracks review count and median review-minutes per reviewer across nights, the first cross-night reviewer signal in the estate (management-api/lib/dream-ledger.js:203-222) |
absent — PR outcome still feeds only a status column, no VisionClaw-owned human-learning measure exists (src/actors/elevation_actor.rs:1131-1200 unchanged from the prior grade) |
| C5 Career pathways (baseline) | measured (baseline) — a reviewer role now exists and an admin's Delegate{to} admits that reviewer's 31403 for exactly the delegated case, giving juniors substantive (not all-or-nothing) review work (crates/nostr-bbs-relay-worker/src/relay_do/nip_handlers.rs:203-249, delegation recorded at :2325) |
absent — no per-reviewer work-allocation mechanism; dream-ledger records who reviewed but does not assign substantive work |
absent — no reviewer-role or delegation mechanism in this substrate; owned by the forum per the field-ownership note below |
| C6 Job purpose (baseline) | measured (baseline) — the published reasoning is the human's own typed text, gated by check_rationale/MIN_HUMAN_RATIONALE_CHARS, and decision_content never fabricates it (crates/nostr-bbs-core/src/governance.rs:711-712,761; crates/nostr-bbs-forum-client/src/utils/governance_view.rs:77); "Human {action} via governance UI" is gone from the codebase (only a removal comment remains, crates/nostr-bbs-forum-client/src/utils/governance_view.rs:624) |
measured (baseline) — a manual continuation requires the operator's own words as evidence, refused when absent (management-api/lib/governance-manual-continue.js:82,105,122-123) |
measured (baseline) — the client rationale gate and server-side check_rationale both require the reviewer's typed text; the fabricated-string path is deleted (client/src/features/control-center/governance/AcspCaseQueue.tsx:16-24; src/handlers/enrichment_proposals_handler.rs:406-423) |
Known limit carried forward from grading, not closed by this merge: the seeded-probe catch rate (feeding forum C4) rides the raw signed 31402, so the measured agent can see its own probe tag and answer it correctly — this is ADR-2011's own review_trigger, tracked as forum docs/security/known-findings.md KF-10 (crates/nostr-bbs-core/src/governance.rs:343,365,596). It does not by itself demote C4 from measured, because the metric exists and runs; it does mean the catch-rate number the metric reports is a floor, not the true rate, until the tagless design in the DDD supersedes it.
Workflow-record field ownership (ADR-2010 §3): agent authority level — agentbox authority_class (agentbox.toml:822-849); review/override/escalation points — forum gates (nip_handlers.rs:715-763); fallback/recovery — agentbox authority gate + forum receipts; unapproved agent actions — agentbox authority-journal (management-api/lib/authority-journal.js:42,92-115 — every authority.deny is a hash-chained record of an action the gate refused to let proceed); effort including verification — forum reviewers telemetry (crates/nostr-bbs-auth-worker/src/governance_api.rs:1287-1288,1336-1394 — median/p90 time-to-decision and override rate per reviewer); effects on human capability — agentbox dream-ledger reviewer columns + forum reviewers telemetry (management-api/lib/dream-ledger.js:203-222; crates/nostr-bbs-auth-worker/src/governance_api.rs:1336-1394 — review-minutes-per-night and calibration/probe catch rate over time are the two cross-cycle capability signals the estate now has). All three prior matrix defects have an owner as of this grading.
Branch state: the three feat/augmentation-conditions branches merged to main — nostr-rust-forum 11b674c, agentbox b859b37d2, VisionClaw 8f9affe49 — closing the milestone tracked in PRD-augmentation-conditions §Milestones (M2). Re-grade on the next substrate-branch merge that touches a cited path, or at the 2026-12-14 longitudinal review trigger, whichever comes first.
The following table is retained as a dated record, not current status. Its replay-only-in-edge, Multikey-canonical-DID, scaffold-only mesh, BrokerActor and twelve-tool claims are superseded by the current table above. Other old counts and verdicts require revalidation before reuse.
| Area | VisionClaw | agentbox | nostr-rust-forum | solid-pod-rs | dreamlab-ai-website | Compatibility status |
|---|---|---|---|---|---|---|
| Identity | NIP-98, NIP-07, DID:Nostr, Solid/WAC documented; enterprise SSO remains a known future path. NIP-26 delegation is not implemented (deferral comments only) | Bootstrap BIP-340 key becomes canonical did:nostr; accepted by relay, pods, WAC, URNs |
Passkey PRF derives Nostr keys; NIP-98 across workers; NIP-42 answering exists only in the WASM browser client, not the edge relay | Foundation for DID:Nostr, NIP-98, Solid-OIDC, Tier 1/Tier 3 DID resolution | WebAuthn PRF + NIP-98; configured agent identities | NIP-98 is reimplemented independently in four repos with divergent URL-matching/replay semantics (G3 open). did:nostr is canonicalised on the Multikey form by ADR-125. NIP-26 delegation is unimplemented everywhere; replay protection exists only in the forum + CF pod tier |
| Mesh | ADR-073 describes client/subscriber mode by default unless a local relay is added | Standalone default ([mesh] mode = "standalone"); embedded relay at :7777 with pod-bridge; federated/client modes available |
nostr-bbs-mesh is scaffold-only (no MeshTransport impl, not a dependency of the relay-worker; lands Sprint v12+). The relay gates by pubkey whitelist, not NIP-42 (auth_required:false). Peer discovery and IS-Envelope routing are not shipped |
Standalone with an embedded NIP-01 relay (no NIP-42 AUTH); "native mesh in alpha.15" is not a real feature — federation is README prose only | Federated architecture; CF Workers relay fan-out to public relays for censorship resistance | Standalone-first is the supported deployment mode (closeout FREEZE). Only dreamlab-ai-website defaults to fan-out — the federated-by-default tally is 1/4, not 2/4 (agentbox, solid-pod-rs and nostr-rust-forum all default standalone). Mesh federation is designed, not shipped |
| Pod | Embedded solid-pod-rs at local :8484; WAC against did:nostr |
Embedded Solid pod and native pod overlay via Cloudflare Tunnel | CF Worker pod tier mirrors Solid/JSS; native tier supports git via solid-pod-rs-server |
Canonical LDP/WAC/WebID/auth/storage implementation | CF pod tier plus native agentbox pod tier for git | Strongest compatibility area; two-tier behavior is the main gap |
| Governance | Merged main runs an inline decide/inbox handler + ADR-110 ElevationActor, not the distributed BrokerActor (that lives on the unmerged crashbug branch only). The POST /api/enrichment-proposals/{id}/decide endpoint performs a real fenced Oxigraph write on attributed approval and flips writeback_committed. Agent Control Surface event kinds defined |
Publishes and subscribes to kinds 31400-31405 via nostr-bridge; 12-tool ontology bridge proxies SPARQL to VisionClaw Oxigraph; handleGovernanceDecision applies returned decisions with PROV-O URNs |
Full implementation of kinds 31400-31405 in governance.rs; primary governance UI/API: agent registry, cases, roles, signed responses |
Access-control foundation, not workflow owner | Governance dashboard at /governance renders kinds 31400-31405; NIP-98 signed responses; feature-gated |
Agent Control Surface implemented in forum + website; agentbox publishes/subscribes and applies decisions. Governance is shipped inline on main (decision application + KG write-back live); the distributed BrokerActor is an unmerged alternative. Remaining wire: ConceptElevated closing event (elevation terminates at PR creation, not merge) |
| Deployment | Docker Compose multi-profile and docs CI; productionisation PRDs still list hardening gaps | Nix-built multi-arch image, runtime contracts, image scan/SBOM, flake checks | Cloudflare Workers kit with CI/audit docs | Server/library deployment docs; CI scope may miss sibling crates | GitHub Pages + five Workers; kit clone/overlay deployment | Usable but split; kit pinning and cross-repo version policy remain weak |
| Ontology Bridge | Oxigraph SPARQL store; OWL 2 EL reasoning (Whelk-rs); 82 CUDA kernels across 9 .cu files (5,854 LOC); canonical knowledge graph |
12 MCP tools (mcp/servers/ontology-bridge.js) proxy SPARQL queries to VisionClaw; gated via [skills.ontology] in agentbox.toml |
Not applicable (forum does not query KG directly) | Not applicable (pod stores data, does not query KG) | Not applicable (website does not query KG) | Shipped cross-substrate integration: agentbox → VisionClaw. Bidirectional (governed propose + query): ontology_propose runs a Whelk EL++ consistency gate then opens a GitHub PR for human review — axiom-submission governance is implemented, not deferred |
| Tests/Ops | Build/test commands and known issues documented; master cross-substrate fixtures relocated to tests/fixtures/ on 2026-06-29 (13 protocol fixtures + 14 schemas — the old docs/specs/fixtures/ path is gone); IS-Envelope spec (ADR-075) |
Upstream fixture vectors under tests/contract/upstream_vectors/ (13 JSON fixtures / 126 vectors / 30 checksummed files) with SHA-256 checksums and upstream pins; adapter contract test suites |
Workspace tests, audit workflow, Worker build docs; fixture copies under tests/fixtures/ — the forum's own CHECKSUMS.txt is red (5 files fail, manifest not regenerated after the 2026-06-15 refresh) |
JSS parity tests/docs; fixture copies in solid-pod-rs-didkey and solid-pod-rs-nostr (did-doc frozen at the pre-refresh state, masked by passing self-checksums) |
README/test docs conflict; some gates advisory | is-envelope-v1.json is byte-identical across all five locations, but did-doc-conformance.json has 4-way drift (7/8/9/7 vectors, two testing a superseded DID spec) — the "checksum verification is the strongest signal" claim does not hold for the security-sensitive DID fixture. sync-fixtures.sh in all three consumers still points at the deleted docs/specs/fixtures/ path |
Every coordinated release should publish:
| Field | Required value |
|---|---|
| Repo SHAs | VisionFlow, VisionClaw, agentbox, solid-pod-rs, nostr-rust-forum, dreamlab-ai-website |
| Identity contract | DID method, public-key form, verification suite string |
| Relay contract | NIP list, write auth, read auth, peer relay list, allowed pubkeys |
| Pod contract | Pod tier, WAC mode, NIP-98 replay behavior, git/export capability |
| Governance contract | Supported Agent Control Surface kinds and schema versions |
| Test evidence | Fixture sync status, smoke test status, audit status |
| Operations | Health endpoints, backup/restore plan, DR owner |
Generate the current local repository SHA set with:
scripts/generate-release-manifest.sh > docs/releases/ecosystem-release.local.jsonSee Release Manifests for the schema and promotion rules.
Historical output, not rerun in this pass. Per ADR-004, each substrate's guide-sensor pairing status is tracked here. This table is generated from scripts/harness-audit.sh (run 2026-07-03) — do not hand-edit; re-run the script and paste its output. See docs/engineering/templates/ for full template definitions.
Two distinct metrics are reported, and they are not the same thing. Paired counts declared guide→sensor cross-references inside a template (a coverage-of-intent measure). Source-backed counts the controls whose source field resolves to a real artifact in the named substrate today; the remainder carry source_status: planned (declared but not yet implemented). A template can be 100% paired while some controls are still planned, so both columns are published to keep the metric honest.
Maturity labels below are the templates' own maturity fields — not hand-typed. governance-decision is a single shared template covering agentbox, nostr-rust-forum and VisionClaw (structure.substrates); it is listed once against all three rather than credited only to the forum.
| Substrate(s) | Harness Template | Guides | Sensors | Paired | Ratio | Source-backed | Maturity |
|---|---|---|---|---|---|---|---|
| agentbox | agentbox-agent-task |
5 | 5 | 5 | 100% | 8/10 (2 planned) | standalone |
| VisionClaw | visionclaw-enrichment |
5 | 5 | 5 | 100% | 8/10 (2 planned) | standalone |
| agentbox + nostr-rust-forum + VisionClaw | governance-decision (shared) |
5 | 5 | 5 | 100% | 9/10 (1 planned) | integrated |
| solid-pod-rs | pod-mutation |
5 | 5 | 5 | 100% | 8/10 (2 planned) | standalone |
| dreamlab-ai-website | — | 0 | 0 | 0 | — | — | planned |
Totals: 20 guides / 20 sensors / 20 paired = 100.0% pairing; 33/40 controls source-backed (7 planned). Audit verdict: PASS (target 80%).
These carry source_status: planned in the templates and are tracked for implementation:
agentbox-agent-task:hooks-validate-trajectory(the ADR-004 D3hooks.validatephase, explicitly deferred),slo-ci-gate(noadapter-slo.yml; the SLO assertions areit.todo()stubs)governance-decision:governance-ui-rendering(no dedicated render test; the governance UI lives in the WASM client)pod-mutation:block-trail-rules+taproot-verification(the solid-pod-rs Taproot implementation DOES exist —crates/solid-pod-rs/src/bitcoin_tx.rsBIP-341/P2TR +provenance.rsBlockTrailAnchor::anchor(); the remaining gap is the VisionFlow-side harness sensor/pairing, not the upstream implementation)visionclaw-enrichment:enrichment-proposal-schema(no standalone JSON Schema fixture yet),provenance-chain-integrity(PROV-O is not yet reified on the ingest path — see closeout T7)
Fitness-gate ADR citations are repo-qualified to avoid ADR-number collisions: agentbox SLO thresholds are agentbox ADR-005 (pluggable-adapter architecture), a different document from VisionFlow ADR-005 (mandate-at-grant governance).
Run scripts/harness-audit.sh for current status.
The area rows above record per-area compatibility posture. Per ADR-002 and the
Gap-Close Canon PRD, each gap-close item is also reflected here at its evidenced
tier once its wave closes. This table is the P0 wave; it is authoritative to
docs/registers/gap-register-v1.1.md (the
immutable status register cut at the P0→P1 boundary). Tiers use the ADR-002
vocabulary; scaffolded here means the mechanism is complete and its canary is
proven locally but has not fired in a live session (a canon loop item is not
scored integrated until its canary fires live — DDD Gap-Close Invariant 4).
| Item | Owner(s) | Closure SHA(s) | Evidenced tier | Canary state |
|---|---|---|---|---|
| RES-a KG liveness + canary harness | VisionClaw | 6f4eb1b0a, 1492bc17b |
integrated |
Harness live; Nostr-tap round-trip PENDING-LIVE |
| RES-b diagram render gate | VisionFlow (canon) | f72d173cd |
scaffolded |
CANARY-CANON-DIAGRAM proven locally; live CI fire PENDING |
| COM-13 agent disclosure | nostr-rust-forum (+ canon norm) | forum 7157a92, fb7826859; canon identity-spine.md |
Forum impl integrated; canon clause standalone |
Badge census-verified; live render PENDING-LIVE |
COM-14 did:nostr keying |
VisionClaw + agentbox | VC 4a595cc8f; AB 6189f47d |
integrated |
Live Schnorr round-trip PENDING-LIVE |
| REC-2 broker kernel + case queue | VisionClaw | c9f2e3539 |
integrated (P0 slice) |
Case e2e PENDING (COM-15 batch) |
| D5 MCP-status honesty | VisionClaw | 6f4eb1b0a |
integrated |
Real WS + check_mcp_metrics; landed |
| REC-1 governed-writeback floor | VisionClaw · solid-pod-rs · forum | 6f4eb1b0a; solid-pod 791977a |
integrated / reconciled |
Route-guard tests green; PATCH + NIP-42 reconciled pre-sprint |
| RES-d drift counter (canon) | VisionFlow (canon) | this wave (P1) | scaffolded |
CANARY-CANON-DRIFT proven locally; live CI fire PENDING |
The forward-chained corrections behind REC-1a/1b, PATCH, NIP-42, D1 and D5 are
recorded in gap-register-v1.1.md §"Forward-chained corrections", not edited into
the v1.0 inventory in place (register immutability, DDD Gap-Close Invariant 5).
The P1 and P2 waves closed the remaining 41 line items. This table reflects them
at their evidenced tiers and is authoritative to
docs/registers/gap-register-v1.2.md (the
immutable status register cut at the P1/P2 boundary, superseding v1.1). Tiers use
the ADR-002 vocabulary. standalone means the mechanism runs within one substrate
but is not yet cross-substrate-proven; scaffolded means the mechanism is complete
and proven locally but its live gate has not fired; a canon loop item is not scored
integrated until its canary fires in a live session (DDD Gap-Close Invariant 4).
The P0 table above is authoritative to v1.1 and is left in place; the two do not
overwrite one another.
| Item | Owner(s) | Closure SHA(s) | Evidenced tier | Canary state |
|---|---|---|---|---|
| F1 member read-only governance view | nostr-rust-forum | 0b8a1c, 6986276 |
integrated |
fired-in-test |
| F3/F7 via COM-16 graduated escalation | nostr-rust-forum | 0b8a1c, 6986276 |
integrated |
fired-in-test |
| F4/F5 via COM-17 decision-audit API | nostr-rust-forum | 0b8a1c, 6986276 |
integrated |
fired-in-test |
| F6 supersession authority | nostr-rust-forum · canon | forum 35dbb1b, 696fd233; canon 7513e58 |
scaffolded (canon-gated) |
fired-in-test |
| F8 roster admin UI | nostr-rust-forum | 0b8a1c, 6986276 |
integrated |
fired-in-test |
| F9 cross-relay federation | VisionFlow (canon) | 7513e58 (fork record) |
planned |
parked (criteria unmet) |
| F10 multi-agent social influence | VisionFlow (canon) | 7513e58 |
standalone (position authored) |
n/a |
| D1 embodiment beam liveness | VisionClaw | 4aca6f729 |
integrated (wired at boot; drift correction) |
PENDING-LIVE (beam traffic) |
| D2/D3/D8 steering + case surface + observability | VisionClaw | e0f582403, 453bd41b1, b6c1c43b5 |
integrated |
fired-in-test; case e2e PENDING-LIVE; MCP-native interrupt a documented boundary |
| D6 PTT directed at selected actor (via COM-15) | VisionClaw · agentbox | f6e1d58f9; AB 9673624, 1fc47a14 |
integrated |
PENDING-LIVE (voice e2e) |
| D7 pre-action intent legibility | VisionClaw · canon | 774ffa05e; canon 2511f994 |
standalone |
fired-in-test |
| M1–M6 / COM-18 headset copresence + intervention | VisionClaw (Godot XR) | 57b32faee, 0f3a1b60c, 348172d7d (merged 485bb8f9f) |
standalone (no headset in box) |
PENDING-LIVE (M4-RAY, COM18-INTERV; M1-HUD rides P0) |
| V1/M5 via COM-15 PTT governed loop | VisionClaw · agentbox | f6e1d58f9; AB 9673624, 1fc47a14 |
integrated |
PENDING-LIVE (voice e2e) |
| V2 multi-agent voice addressing | VisionFlow (canon) | 7513e58 |
standalone (position authored) |
n/a |
| V3 conversational grounding/repair | VisionClaw | 774ffa05e |
standalone |
fired-in-test |
| V4 voice-docs honesty | VisionClaw | 774ffa05e |
integrated (docs) |
n/a |
| REC-2 broker case queue (P1 slice) | VisionClaw | e0f582403 |
integrated |
case e2e PENDING-LIVE |
| REC-3 contextual transaction cost | VisionClaw · agentbox | 4aca6f729, 453bd41b1; AB 9673624, ceb3401b |
integrated |
fired-in-test |
| REC-4 four-KPI dashboard (ADR-043) | VisionClaw | 4aca6f729 |
integrated (2 KPIs live; 2 honest pending tiles) |
fired-in-test |
| REC-5 MAST failure telemetry | agentbox · VisionClaw | AB 9673624, 1fc47a14, ceb3401b |
integrated |
fired-in-test |
| REC-6 escalation-default authority | agentbox | 9673624, 1fc47a14, ceb3401b |
standalone (re-tiered honestly) |
fired-in-test |
| REC-7 outcome learning | agentbox | 9673624, 1fc47a14 |
standalone (consumers gated pending floor) |
fired-in-test |
| REC-8 orchestration diversity | agentbox | 9ebff750, a8dd21a5 |
standalone (re-tiered) |
fired-in-test |
| REC-9 provenance-to-pocket | agentbox | 9ebff750, a8dd21a5 |
standalone (re-tiered) |
fired-in-test |
| REC-10 Insight Ingestion Loop v1 | VisionClaw · agentbox · forum | 1c462f492 |
integrated |
fired-in-test |
| REC-11 data-moat unified trace | VisionClaw · solid-pod-rs | 1c462f492; SP 40043b0 (+WP-2) |
standalone (pod _prov/ contract-only) |
fired-in-test |
| REC-12 kit cutover + external pilot | dreamlab-ai-website | ca06ab3 |
standalone (pilot blocked-operational) |
registered (D3 key-split runbook) |
| COM-15 PTT voice-to-actor loop | VisionClaw · agentbox | f6e1d58f9; AB 9673624, 1fc47a14 |
integrated |
PENDING-LIVE (voice e2e) |
| COM-16 graduated escalation + fatigue | nostr-rust-forum | 0b8a1c, 6986276 |
integrated |
fired-in-test |
| COM-17 decision-audit + trust-calibration | nostr-rust-forum | 0b8a1c, 6986276 |
integrated |
fired-in-test |
| COM-18 headset intervention + identity | VisionClaw (Godot XR) | 57b32faee, 0f3a1b60c, 348172d7d |
standalone (no headset) |
PENDING-LIVE (COM18-INTERV) |
| RES-c solid-pod-rs diagram refresh | solid-pod-rs | 40043b0 (+WP-2) |
integrated (9 diagrams re-rendered) |
fired-in-test |
| RES-d self-description drift counter + CI | VisionFlow (canon) | b47c5fd; sources VC 4aca6f729, AB d13f8688 |
scaffolded |
PENDING-LIVE (CANARY-CANON-DRIFT CI fire) |
| RES-e Wardley export quality | VisionFlow (canon) | 7513e58 |
integrated |
n/a |
The P1/P2 forward-chained corrections (REC-2/ADR-041 "Implemented" claim, the D2
MCP-native interrupt boundary, and the REC-6/REC-8/REC-9/F6 tier re-derivations)
are recorded in gap-register-v1.2.md §"Forward-chained corrections", not edited
into v1.1 or the v1.0 inventory in place. The pending-live canaries are batched to
one stack-up session at sprint end (a recorded deviation from strict wave gating,
because no live stack, headset or GitHub Actions push runs in the build
container); see gap-register-v1.2.md §"Pending-live-session batch".