You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
A committed-tree review of origin/main...origin/dev found several documentation statements that lag the implemented and landed contracts. The only runtime artifact affected is static alert annotation text; alert evaluation remains unchanged.
Required corrections
ADR-0012: state the exact per-field per-fact OpenCost magnitude, the 4096-fact aggregate bound, the derived aggregate maximum, and that all fifteen cost fields are accumulated independently.
Worker-pool session: repair the duplicated wording for the one-second parallel timeout wave.
Kubeconfig race session: distinguish rejected absolute or relative executable paths from supported bare commands resolved through PATH.
E2 integration spec: preserve the existing connector-wave scope and describe the Wave-1 core R6 boundary as detection plus Prometheus quantification; cloud-autoscaler facts remain required for confidence.
README audit paging: distinguish online per-page database and response-egress cost from offline local verification, which performs no network I/O.
Architecture model: include immutable approval expiry and the exact half-open consumption interval.
F10.4g alert documentation and static annotation: say that the freshness guard proves a recent scraped sample from the preinitialized accepted series, not an accepted authentication event.
Acceptance criteria
Every changed statement matches current implementation, tests, and live merge evidence.
No alert firing expression, roadmap priority, connector-wave scope, IAM, cloud resource, or recurring-cost change.
Markdown and repository contract checks pass.
Independent exact-diff review reports no unresolved finding.
Explicit non-findings
ADR-0012 stale-input rejection already exists in the generic graph fact contract and rollup tests.
The chart already documents the browser OIDC session-private.pem requirement.
The alert suite already proves that one accepted authentication suppresses refusal-only traffic and that absent or stale accepted telemetry stays quiet.
Security and cost
This is evidence-contract repair plus a static annotation precision fix. It does not widen IAM, networking, secret access, telemetry cardinality, egress, storage, or recurring cost.
Problem
A committed-tree review of
origin/main...origin/devfound several documentation statements that lag the implemented and landed contracts. The only runtime artifact affected is static alert annotation text; alert evaluation remains unchanged.Required corrections
Acceptance criteria
Explicit non-findings
session-private.pemrequirement.Security and cost
This is evidence-contract repair plus a static annotation precision fix. It does not widen IAM, networking, secret access, telemetry cardinality, egress, storage, or recurring cost.