Warden · acceptance & criteria evals

Robin Evaluation Suite

61 plans · 3 executable criteria evals · 25 pending · 33 v1-era acceptance plans

Status. The criteria register is pending owner acceptance — no criterion is binding until graded and accepted. The three executable plans cannot run until the warden environment is provisioned on this machine (env provisioning currently exits 3). The 28-series plans report as pending by design: each names the architecture precondition it waits on.

Companion docs: criteria register · tier definitions

Criteria evals — runnable now (3)

PlanCriterionStatus
25 - CR-18: Domain classifier quality (Signal→Domain macro-F1)CR-18executable
26 - CR-25: ALCE citation-fidelity floor (current pipeline)CR-25executable
27 - CR-20: Entry→Signal extraction completenessCR-20executable

Criteria evals — pending architecture (25)

PlanCriterionStatus
28 - CR-01 (pending): Dimension-aware reranking gainCR-01pending
28 - CR-02 (pending): Standing penalty gainCR-02pending
28 - CR-03 (pending): Claims-in-index retrieval contributionCR-03pending
28 - CR-04 (pending): Claims-layer artifact qualityCR-04pending
28 - CR-05 (pending): Signal-branch contradiction detectionCR-05pending
28 - CR-06 (pending): Claim-branch position filterCR-06pending
28 - CR-07 (pending): Retraction minimal-change checkCR-07pending
28 - CR-08 (pending): Initiative attention gainCR-08pending
28 - CR-09 (pending): Domain scoping gainCR-09pending
28 - CR-11 (pending): Visibility rule decisionCR-11pending
28 - CR-12 (pending): Fast/slow compoundingCR-12pending
28 - CR-13 (pending): Staleness under injected contradictionsCR-13pending
28 - CR-14 (pending): Cost ledger break-evenCR-14pending
28 - CR-15 (pending): Dimension meaningfulnessCR-15pending
28 - CR-16 (pending): Small-projector agreementCR-16pending
28 - CR-17 (pending): Router accuracyCR-17pending
28 - CR-19 (pending): Anchor thesis falsification (composite)CR-19pending
28 - CR-21 (pending): Stance-conflict exclusion: contaminationCR-21pending
28 - CR-22 (pending): Revision succession: survival and removal retractionCR-22pending
28 - CR-23 (pending): Stance boundary audit: leak rate, shadow run, classifier fallbackCR-23pending
28 - CR-24 (pending): Artifact engine: stance legibility with citation fidelityCR-24pending
28 - CR-26 (pending): Live-artifact staleness and stamp coverageCR-26pending
28 - CR-27 (pending): Label-free eviction: engagement as salience proxyCR-27pending
28 - CR-28 (pending): The dial: auto-activation qualityCR-28pending
28 - CR-29 (pending): Companion thesis falsification (composite)CR-29pending

V1 / V1.1 acceptance plans (33)

PlanCriterionStatus
00 - Warden's own gates, inside warden's suite—shipped era
01 - Greenlight quality gates—shipped era
02 - Foundation substrate invariants—shipped era
03 - Connect interaction-surface invariants—shipped era
04 - Permissions authorization-engine invariants—shipped era
05 - Collapse domain≡signal-database invariants (D34)—shipped era
06 - Socrates DANCE end-to-end acceptance (D33, W1–W7)—shipped era
07 - wiki_types workspace scoping (D35)—shipped era
08 - workspace reads on the ability engine (D36)—shipped era
09 - root-workspace internal-only gate (D30)—shipped era
10 - app shell acceptance (session E2E, IA reachability, tokens, rendered rails)—shipped era
11 - capture gate (audio ingestion: PKCE client → upload → transcription)—shipped era
12 - ask-edit gate (D38 suggestion intelligence: ask loop, auto-accept, inbox ranking)—shipped era
13 - bug-sweep (workspace & authoring correctness)—shipped era
14 - member-wiki authoring & promotion governance lifecycle—shipped era
15 - v1.2 promotion payload (better-auth 1.7 migration + #136 + agent hardening)—shipped era
16 - Corpus evals over the seed fixture—shipped era
17 - workspace/org content isolation across screens + MCP tools (#158)—shipped era
17 - issue #187 MCP connector UX (credential-less discovery, plain endpoint, branded metadata)—shipped era
17 - issue #198: passcode retired from the user-facing UI—shipped era
17 - wiki-type ghost retirement (issue #306, batch v11-batch)—shipped era
17 - domain-wiki detach guard (issue #349)—shipped era
18 - issue #337 entity/client domain inheritance for entity-scoped signals—shipped era
18 - admin control plane IA: People & Access / Content / Organization (issue #214)—shipped era
18 - composite child-change dirty cascade + Repair re-weave (issue #224)—shipped era
19 - issue #187 MCP connector UX: credential-less discovery, plain endpoint, branded metadata—shipped era
20 - first-author grounding (issue #328)—shipped era
20 - the inbox counts decisions, not timestamps (issue #347)—shipped era
21 - issue #366 provenance repair (the #338 nudge carries real inputs)—shipped era
21 - a user-created wiki type generates from a neutral base, not POV (issue #367)—shipped era
22 - ownership, org transfer, agent gaps, error visibility (issue #365)—shipped era
23 - MCP search scope is membership-gated (issue #387)—shipped era
24 - The root workspace is visible only to its own members (issue #391)—shipped era