[mnkiefer]

Summary

• Show reviewed Value registrations and authoritative collection health before a first report, both on the Value page and workflow-level surface. Distinguish disabled, unconfigured, never captured, deferred (including actual API-budget reasons), failed, and retained evidence without inventing freshness.
• Carry separate collection facts through the real Activity producer, publication, normalization, canonical worker queries, browser/SQLite storage, and hosted PostgreSQL paths. Decision time, actual attempt, accepted capture and evidence cutoff remain distinct; missing evidence is not zero.
• Remove automatic remote Value acquisition from ordinary Activity/hosted projection. Preserve accepted packets and proof, and run eligible deterministic local-only computation independently with reviewed adapter admission, separate concurrency/leases, credential exclusion, time/output limits, and conflict-safe result import.
• Apply the useful narrow part of the proposed storage formalism: replay-order-independent decision/attempt reconciliation, same-clock conflict rejection, and hosted publication guards against changed generation, policy, manifest or accepted evidence. This is not a three-tier storage implementation, durable evidence platform, or incremental scheduler.

Capacity and authority boundaries

Recurring remote acquisition remains deferred pending enforceable shared-budget admission. This slice improves health visibility and core isolation, not guaranteed fresh Value collection. Enabled registrations remain visibly registered-but-deferred, not disabled. A separate worker is not separate API quota; reviewed local-only source admission is not an OS network sandbox.

Bounded cache-only/local deterministic computations remain supported. Existing exact opt-in, explicit disables, public-only/fast/API-limited gates and frozen measurement contracts are preserved. No forced acquisition, automatic broad backfill, target activation, scope expansion, new credentials, production dispatch, deployment or merge.

.github/workflows/cao.json is unchanged, including daily-repo-status: enabled:false and dashboard deploy:false. Publication ownership remains Activity → Dashboard artifact/cache → Documentation Pages /cao/. Retained evidence keeps its actual capture/cutoff when newer collection is deferred or fails.

Validation

Check	Result
Full npm run check	Passed: 1,791 unit tests (1 skipped), 321 integration tests, 10 load tests, 29 schema tests, lint/typechecks/SVG checks, 74 workflow compilations and documentation build
Dashboard unit/E2E	Full suites passed: 3,714 unit tests (1 skipped), 300 E2E tests; final focused persistence/view tests 87 passed and canonical/Value E2E 26 passed
Dashboard performance/memory	All 24 Lighthouse scenarios passed (performance 97–100, maximum CLS 0.014); final deployed-data worker-memory gate passed with unchanged budgets
Hosted parity/isolation	Live PostgreSQL/Redis -race -p 1 suites passed for postgresx, collect, operationalvalue, sqlitex and ingest; Go lint passed
Contracts and distribution	Canonical query-tool evaluation, producer-to-storage roundtrip, installed-runtime smoke, publication/resource bounds, timeout/process cancellation, six replay permutations and superseded-input guards passed
Mergeability	Nonmutating merge check against current origin/main passed

The full gate used a task-local gh-aw v0.93.1, matching the reviewed policy rather than changing the machine-wide compiler. Local Node is v25; NODE_NO_WARNINGS=1 suppresses its SQLite experimental-warning stderr for existing exact-stderr assertions. Compiler output retains two existing warnings and action-version notices; compilation is not live workflow execution. Dashboard full-suite/performance results preceded the final narrow reconciliation changes, which were covered again by focused storage/E2E and memory checks.

Desktop/mobile previews use the real built full site and production producer/ingestion/worker path with explicitly synthetic fixtures. They are UI review evidence, not fresh GitHub measurements; screenshots are attached in a PR comment and are not committed.

Read-only boundary review and authored documentation changes

PASS: one read-only reviewer checked intent/effect scope, credential destinations and disclosure, executable provenance, persistence/cleanup, untrusted inputs, authorization and preserved controls. All ten questions received source-backed YES answers. No confirmed vulnerabilities were found in the reviewed changes; this is not a live deployment attestation. Reviewer model identity was unavailable.

#	Severity	File	Lines	Vulnerability	Confidence

No authored workflow Markdown or generated lock YAML changed. Conventional Actions YAML and the following directly related authored documentation did change:

Markdown source	Change	Security impact
Operational Value skill:232–242	Automatic-refresh guidance becomes enabled-but-deferred registration.	Does not grant quota or new authority; reviewed local-only adapters remain bounded.
Architecture:737–760	Inline collection becomes separate health facts and guarded local publication.	Preserves independent clocks and core isolation.
Activity runtime:345–369	Documents independent local work, eligibility gates and retained results.	Core publication does not await optional computation.
CLI:497–520	Registration no longer promises acquisition.	Preserves explicit disables and separates registration from activation.
Ingestion:184–200	Documents schema versions and collection-health storage.	Derived storage adds no collection authority.
Data model:123–130	Adds separate repository/workflow health facts.	Prevents health timestamps from becoming outcome captures.
Value how-to:113–123	Recurring-refresh promise becomes budget-dependent remote deferral.	Removes an unsupported freshness guarantee.
Value contract:84–120	Distinguishes registration, health, local work and retained evidence.	Missing reports do not infer configuration or change deployment ownership.
Hosted runtime:2858–2880	Credentialed recurring acquisition becomes health plus separately leased local computation.	Documents resource limits and publication preconditions.
Normative data contract:2935–3029	Defines current-fact health, replay reconciliation and superseded-input rejection.	Explicitly excludes broader storage infrastructure and unsupported isolation claims.

[mnkiefer]

Synthetic full-site previews — desktop (1440px) and mobile (390px). Captured from the built dashboard served by Go, after the real producer, normalization and worker-query path rendered. The fixtures show an enabled API-deferred registration before its first capture, a disabled registration, and retained evidence with its original dates. These are UI previews, not fresh GitHub measurements; no production collection or deployment was performed.

[github-actions]

Dashboard dead views

Dead element views (0):
Dead dashboard pages (0):
Analyzed 8 element renderers and 68 dashboard pages across 1 dashboard.json file(s) from dashboard/site/src/components/ui-elements.js.

⚠️ This reachability report is informational and does not block the pull request.
🔎 View the workflow run and report artifact.

[mnkiefer]

@copilot fix failing checks

[Copilot]

@copilot fix failing checks

Fixed the native query parity failure by registering workflowValueCollections in the native snapshot allowlist. The targeted Go parity test passes, and the dashboard storage-shards test that timed out in CI passes locally. Commit: 2efffc4. The new PR workflow runs are currently action_required.