Files
canina/specs/002-seo-performance-tracking/acceptance-matrix.md
T

7.0 KiB

Independent acceptance matrix

Audit date: 2026-10-28 UTC. Status OPEN means not independently exercised in this read-only PMO audit, not an assertion that concurrent implementation lacks it. Bind receipts to the final integrated SHA; identify environment, actual command/request, sanitized response, assertion, exit code and artifact path. Store planned execution evidence under evidence/seo-tracking/<SHA>/<row-id>/; this is a naming convention, not a claim those artifacts exist. This audit ran Spec Kit prerequisites and read-only transport/source checks only; no application acceptance is claimed.

ID / requirement Owner(s) Concrete scenario and expected assertion Required independent evidence Status
Q01 / R2 auth Engineering + Security/QA GET list/detail and POST finding/snapshot, PATCH feedback/action with no credential →401; real non-admin login →403; authorized login succeeds; no guard bypass Real HTTP responses from isolated app and fresh approved-vault Stage login (not forged live JWT) OPEN
Q02 / R2 input Engineering/DBA Invalid UUID, extra fields, oversized/negative page, unapproved sort, NaN/negative metrics, inverted/future window, foreign finding/target are rejected without writes HTTP validation and before/after real DB assertions OPEN
Q03 / R1 multiplicity Research/SEO + Engineering Persist one finding with ≥2 actions/targets, ≥3 keywords and ≥2 independently defined goals; query all after process/database client restart; deleting target preserves evidence/history Actual API write/read isolated DB, persisted relations/history, goal units/baselines OPEN
Q04 / R2 query Backend + Frontend Two pages with equal sort values have no duplicates/missing IDs; real total consistent; Persian/Arabic variants search correctly; filter/search resets page and uses server state API + browser request/DOM assertions and stable ID set comparison OPEN
Q05 / R3 feedback race Backend + QA Two saves with same expected version: only one succeeds, other 409; later GET has winning text/actor/time/version; no silent overwrite on UI retry Concurrent real HTTP/DB test and browser conflict/reload interaction OPEN
Q06 / R3 downstream enforcement Content/Medical + SEO + Engineering Capture current directive, begin actual editorial/SEO action, change feedback before finalize; stale action is blocked; retry rereads directive and action evidence identifies compliance/version Real consumer integration logs/artifact and absence of stale side effect; controller-only test insufficient OPEN
Q07 / R4 nullable semantics Worker + Backend + UI Real zero-count provider window displays 0; 401/403/timeout/unsupported/missing metric displays null + reason/time, never 0; rank without observations is not zero; one missing metric does not falsify others Sanitized provider response→HTTP→DB→DOM trace; deterministic error unit test supplement OPEN
Q08 / R6 idempotence Engineering/DBA + Worker Concurrent identical ingestion and retry after restart create exactly one snapshot; changed payload same key →409; new window appends; old data unchanged Real PostgreSQL uniqueness/race/restart test and exact count/ID comparison OPEN
Q09 / R6 history Backend + UI More than 30 windows retrievable with stable pagination/explicit bounds; source window/canonical URL/units/time preserved; no misleading chart continuity across missing periods API history and UI inspection against actual DB set OPEN
Q10 / R4 Google Research/SEO + DevOps Real authorized GSC domain/GA4 property read for linked canonical URL/window; reconcile counts/aggregates/date timezone; test canonical aliases/provider pagination/quota errors Actual sanitized API results, query dimensions and comparison receipt; mocks do not pass live row OPEN
Q11 / R4 Clarity Frontend/UX + DevOps Discover approved report access and fetch/import real behavior report with provenance; if unavailable, UI identifies exact limitation and owner, no made-up Clarity scores Actual report/request or explicit unsupported receipt; script presence alone insufficient OPEN
Q12 / R4 event goals Growth + Frontend/UX Subscription success vs attempt, add_to_cart quantity/value, checkout, article→product/shop/internal navigation and purchase events verified independently; no duplicate route event; scroll threshold/denominator and engagement/bounce units correct Browser network/collector payload plus real reporting receipt by URL/window; missing instrumentation recorded unavailable OPEN
Q13 / R5 commerce Growth/Commerce + DBA + QA Prove lawful article/session/order join; paid/completed/refunded/deduped orders and currency/window/model reconcile; unknown join/cross-device/consent loss never maps total orders to article or zero attributable revenue Documented mechanism and sanitized real aggregate reconciliation; explicit unknown limitations if unsupported OPEN
Q14 / R2 UI Frontend/UX + QA Real sidebar route loads data; loading/empty/error/retry, nullable sources/limitations/history/goals/directive conflicts visible; mobile viewport/keyboard and RTL inputs usable Authenticated Stage DOM/network interaction evidence on actual built bundle OPEN
Q15 / R7 alerts DevOps + department owners + PMO Provider failure creates routed deduped department work/alert, failed delivery persists/retries, recovery recorded; actual receiving department receipt, no raw secret/error dump Real approved delivery and receiving-side readback; mocked Telegram send insufficient OPEN
Q16 / R6 decisions SEO + Medical + PMO Sparse/missing/comparable-vs-incomparable windows yield uncertainty; arbitrary thresholds cannot prune/delete/change content or force PRUNED; recommendation remains distinct from task authorization/execution Tests plus task/decision audit; no destructive live probe OPEN
Q17 / constraints schema/release DBA + DevOps + QA Reviewed schema compatibility/separate authorized migration; full suites/builds pass final SHA; exact-SHA Actions terminal log, both image/container labels and live feature behavior agree DDL/backup approval, actual test exits, Actions URL/log, manifest/images and Stage acceptance OPEN
Q18 / R7 scheduling DevOps + Parsa Deliver tested invocable worker with documented dry-run mutation boundary; no cron created/activated by agents; Parsa configures jobs individually Worker CLI run and read-only scheduler inventory when available; explicit ownership note OPEN

Omissions corrected by this audit

Original spec collapsed research to one blog and one expected-outcome string, omitted many-action/goal history and Clarity acceptance, treated all metric defaults as zero, asserted commerce attribution without a join, limited anomalies to Telegram, omitted stale directive enforcement and durable idempotence, and instructed automatic cron creation. Those are acceptance obligations now; edits do not prove implementation. No arbitrary prune/scaling threshold is authorized.