{"format": "ta-amps.seed-pack.verbs.v1", "pack_id": "self-rewarding-and-iterative-dpo-flywheel", "status": "proposal_only", "tasks": [
{"queue": "conformance-seed-packs", "kind": "verb.probe_source", "effect_class": "write-internal", "priority": 50, "idempotency_key": "seedpack:self-rewarding-and-iterative-dpo-flywheel:calibrate-rubric-judge-against-verifiers:bebb5071d735a0e617e569f6551c4dfd515faa817990f8dc0219db91b5c315ea", "depends_on": [], "producer_surface": "surface:tessera", "visibility_scope": "surface_only", "allowed_surfaces": [], "target_cohorts": ["all"], "tags": ["category:verb.probe_source", "pack:self-rewarding-and-iterative-dpo-flywheel", "template:calibrate-rubric-judge-against-verifiers", "claimant:agent", "status:quarantined"], "brief": {"category_id": "verb.probe_source", "contract": "ta-amps.seed-verb-template.v1", "pack": "self-rewarding-and-iterative-dpo-flywheel", "pack_version": "0.1.0", "template_id": "self-rewarding-and-iterative-dpo-flywheel.calibrate-rubric-judge-against-verifiers", "title": "Calibrate candidate self-rewarding rubric judges against deterministic ground-truth verifiers", "plain_summary": "Evaluate rubric judge approvals against deterministic verifier outcomes using Cohen's kappa, exact agreement and false-positive approval rate bounds, blocking synthetic pairs when calibration fails.", "outcome": "Rubric judge calibration scorecard and admitted/blocked synthetic pair counts.", "claimant": "agent", "inputs": {"pinned": [{"store": "seedpack.self-rewarding-and-iterative-dpo-flywheel", "key": "self-rewarding-and-iterative-dpo-flywheel.v0.1.0.spec.iterative-rpo-and-flywheel-grid", "version": 1, "value_sha256": "7718341ff5d936823077ad1990255647603f58735bc00d10503a9c4606b14460"}, {"store": "seedpack.self-rewarding-and-iterative-dpo-flywheel", "key": "self-rewarding-and-iterative-dpo-flywheel.v0.1.0.rules.rpo-objective-and-calibration-gates", "version": 1, "value_sha256": "22235e9bb2d7131cafbcd0b9ecb917cb86cb9104d0d3386c359ff45539c90286"}, {"store": "seedpack.self-rewarding-and-iterative-dpo-flywheel", "key": "self-rewarding-and-iterative-dpo-flywheel.v0.1.0.guardrail.block-uncalibrated-synthetic-rubric-pairs", "version": 1, "value_sha256": "7cb46a0edcc6becf59dc750a792ab8e46ba519b613cf9ee2015c7b54e24df548"}, {"store": "seedpack.self-rewarding-and-iterative-dpo-flywheel", "key": "self-rewarding-and-iterative-dpo-flywheel.v0.1.0.tool.iterative-rpo-flywheel", "version": 1, "value_sha256": "4be9841095679303522ff8e2f28440424fe39b339a5fdf7461a5beb724042130"}], "slots": {"calibration_bundle": null}, "params": {"output_store": "seedpack.self-rewarding-and-iterative-dpo-flywheel.outputs"}}, "slot_specs": {"calibration_bundle": {"category": "noun.eval_run", "required": true, "many": false, "description": "Pin of paired rubric judge predictions, deterministic verifier verdicts and candidate synthetic pairs."}}, "param_specs": {"output_store": {"description": "Store for the produced nouns. The default is a placeholder; use the store your owner assigned.", "schema": {"type": "string", "pattern": "^[a-z0-9][a-z0-9_.-]{0,63}$"}, "required": true, "default": "seedpack.self-rewarding-and-iterative-dpo-flywheel.outputs"}}, "input_schema": {"type": "object", "additionalProperties": false, "required": ["pinned", "slots", "params"], "properties": {"pinned": {"type": "array", "minItems": 4, "maxItems": 4, "items": {"type": "object", "additionalProperties": false, "required": ["store", "key", "version", "value_sha256"], "properties": {"store": {"type": "string", "pattern": "^[a-z0-9][a-z0-9_.-]{0,63}$"}, "key": {"type": "string", "pattern": "^[A-Za-z0-9][A-Za-z0-9_.:-]{0,255}$"}, "version": {"type": "integer", "minimum": 1}, "value_sha256": {"type": "string", "pattern": "^[0-9a-f]{64}$"}}}}, "slots": {"type": "object", "additionalProperties": false, "required": ["calibration_bundle"], "properties": {"calibration_bundle": {"type": "object", "additionalProperties": false, "required": ["store", "key", "version", "value_sha256"], "properties": {"store": {"type": "string", "pattern": "^[a-z0-9][a-z0-9_.-]{0,63}$"}, "key": {"type": "string", "pattern": "^[A-Za-z0-9][A-Za-z0-9_.:-]{0,255}$"}, "version": {"type": "integer", "minimum": 1}, "value_sha256": {"type": "string", "pattern": "^[0-9a-f]{64}$"}}}}}, "params": {"type": "object", "additionalProperties": false, "required": ["output_store"], "properties": {"output_store": {"type": "string", "pattern": "^[a-z0-9][a-z0-9_.-]{0,63}$"}}}}}, "output_contract": {"visibility_scope": "surface_only", "allowed_surfaces": [], "write_mode": "create_only", "store_param": "output_store", "key_rule": "self-rewarding-and-iterative-dpo-flywheel.<role>.<task_id>[.<n>]", "expected_version": 0, "produces": [{"name": "calibration_receipt", "template": {"store": "seedpack.self-rewarding-and-iterative-dpo-flywheel", "key": "self-rewarding-and-iterative-dpo-flywheel.v0.1.0.template.iterative-rpo-flywheel-receipt", "version": 1, "value_sha256": "a1f71af1350671ca8cb681c98d855404a3bdcf5bb5cec750dbabb885c621d280"}, "max": 1}], "delivery": "Write produced nouns surface_only under your own authenticated surface, which is the requester's assigned surface, so the requester and that surface's reviewers can read them. Never use public_commons or allowlist other surfaces. Return every produced noun's pin in output.artifacts; add its value when the requester asks for inline delivery."}, "output_schema": {"type": "object", "additionalProperties": false, "required": ["outcome", "summary", "artifacts", "checks", "unknowns"], "properties": {"outcome": {"type": "string", "enum": ["rubric_judge_calibrated", "no_change", "unknown", "blocked"]}, "summary": {"type": "string", "minLength": 1, "maxLength": 1200}, "artifacts": {"type": "array", "maxItems": 1, "items": {"type": "object", "additionalProperties": false, "required": ["role", "ref"], "properties": {"role": {"type": "string", "minLength": 1, "maxLength": 64}, "ref": {"type": "object", "additionalProperties": false, "required": ["store", "key", "version", "value_sha256"], "properties": {"store": {"type": "string", "pattern": "^[a-z0-9][a-z0-9_.-]{0,63}$"}, "key": {"type": "string", "pattern": "^[A-Za-z0-9][A-Za-z0-9_.:-]{0,255}$"}, "version": {"type": "integer", "minimum": 1}, "value_sha256": {"type": "string", "pattern": "^[0-9a-f]{64}$"}}}, "value": {"type": "object"}}}}, "checks": {"type": "array", "maxItems": 64, "items": {"type": "object", "additionalProperties": false, "required": ["id", "result"], "properties": {"id": {"type": "string", "minLength": 1, "maxLength": 80}, "result": {"type": "string", "enum": ["pass", "fail", "not_run"]}, "evidence": {"type": "string", "maxLength": 2000}}}}, "unknowns": {"type": "array", "maxItems": 32, "items": {"type": "object", "additionalProperties": false, "required": ["item", "reason"], "properties": {"item": {"type": "string", "minLength": 1, "maxLength": 300}, "reason": {"type": "string", "minLength": 1, "maxLength": 1000}, "would_resolve": {"type": "string", "maxLength": 1000}}}}, "no_change_reason": {"type": "string", "minLength": 1, "maxLength": 2000}, "blocked_question": {"type": "string", "minLength": 1, "maxLength": 1000}, "proposed_verbs": {"type": "array", "maxItems": 10, "items": {"type": "object", "additionalProperties": false, "required": ["kind", "title", "outcome", "why", "priority"], "properties": {"kind": {"type": "string", "pattern": "^verb\\.[a-z_]+$"}, "template_id": {"type": "string", "minLength": 1, "maxLength": 160}, "title": {"type": "string", "minLength": 1, "maxLength": 160}, "outcome": {"type": "string", "minLength": 1, "maxLength": 600}, "why": {"type": "string", "minLength": 1, "maxLength": 600}, "input_pins": {"type": "array", "maxItems": 16, "items": {"type": "object", "additionalProperties": false, "required": ["store", "key", "version", "value_sha256"], "properties": {"store": {"type": "string", "pattern": "^[a-z0-9][a-z0-9_.-]{0,63}$"}, "key": {"type": "string", "pattern": "^[A-Za-z0-9][A-Za-z0-9_.:-]{0,255}$"}, "version": {"type": "integer", "minimum": 1}, "value_sha256": {"type": "string", "pattern": "^[0-9a-f]{64}$"}}}}, "priority": {"type": "integer", "minimum": 1, "maximum": 3}}}}, "proposed_nouns": {"type": "array", "maxItems": 10, "items": {"type": "object", "additionalProperties": false, "required": ["category", "role", "title", "why"], "properties": {"category": {"type": "string", "pattern": "^noun\\.[a-z_]+$"}, "role": {"type": "string", "minLength": 1, "maxLength": 64}, "title": {"type": "string", "minLength": 1, "maxLength": 160}, "why": {"type": "string", "minLength": 1, "maxLength": 600}}}}, "run_notes": {"type": "object", "additionalProperties": false, "properties": {"went_right": {"type": "array", "maxItems": 8, "items": {"type": "string", "minLength": 1, "maxLength": 500}}, "went_wrong": {"type": "array", "maxItems": 8, "items": {"type": "string", "minLength": 1, "maxLength": 500}}, "next_steps": {"type": "array", "maxItems": 8, "items": {"type": "string", "minLength": 1, "maxLength": 500}}, "ideas": {"type": "array", "maxItems": 8, "items": {"type": "string", "minLength": 1, "maxLength": 500}}}}}}, "permitted_outcomes": ["rubric_judge_calibrated", "no_change", "unknown", "blocked"], "outcome_meanings": {"rubric_judge_calibrated": "The declared nouns were produced and every deterministic acceptance check passed.", "no_change": "The evidence does not justify producing or changing anything. Say why.", "unknown": "The question could not be answered with the permitted inputs. List what is missing and what would resolve it.", "blocked": "A required input, permission or capability is missing. Ask one precise question."}, "outcome_requirements": {"rubric_judge_calibrated": ["artifacts"], "no_change": ["no_change_reason"], "unknown": ["unknowns"], "blocked": ["blocked_question"]}, "acceptance_checks": [{"id": "uncalibrated-judges-blocked", "check": "Uncalibrated rubric judges admit zero synthetic pairs while calibrated judges satisfy kappa and false-positive bounds.", "method": "deterministic"}], "creates": {"nouns": "The worker writes the nouns listed in output_contract.produces and returns their pins.", "verbs": "The worker proposes follow-up verbs in output.proposed_verbs, such as a breakdown into smaller verbs or a research follow-up, with the pins they need. Proposals are never enqueued by the worker; an authorized requester enqueues the ones worth doing."}, "guidance": {"required": ["Load Cohen's kappa, exact agreement and false-positive rate bounds from default_grid_spec.json.", "Evaluate candidate rubric judges on deterministic verifier labels and block synthetic pairs from uncalibrated judges.", "Return follow-up work the results justify as proposed_verbs and nouns worth creating later as proposed_nouns. Never enqueue them yourself; an authorized requester decides which to run."], "preferred": [], "suggested": [], "discovery_requested": false}, "required_capabilities": ["read permitted noun versions", "write candidate nouns in the requester's scope"], "bounds": {"max_attempts": 2, "max_elapsed_seconds": 1800}, "execution": {"reasoning": "deterministic_tools", "model_use": "none", "inference_allowance": "Existing authorized worker allowance only. This template grants no provider credit, spending authority or server inference.", "server_fallback": false, "maximum_automatically_enqueued_children": 0, "network": {"mode": "none", "detail": "No network access."}, "external_effects": "none"}, "review": {"required": "owning_policy", "self_approval_permitted": false, "publication_permitted": false, "admission_permitted": false}, "instantiation": {"status": "template", "fill_before_enqueue": ["calibration_bundle"], "rule": "Replace each slot with exact NounRef pins, set the params, then recompute the idempotency key with scripts/build_packs.py --instantiate. Enqueue only with your own grants in a queue your owner assigned."}, "execution_report": {"optional": true, "purpose": "Lets later runs of this template be fingerprinted and compared. It is worker reported, not proof.", "template": {"version": 1, "harness": {"name": "replace-with-harness-name", "version": "replace-with-exact-version"}, "model": {"provider": "replace-with-provider-or-none", "name": "replace-with-exact-model-or-none"}, "summaries": {"problem": "One sentence: what this run was asked to resolve.", "input": "One sentence: which pinned inputs were read.", "solution": "One sentence: what was produced or why the outcome was no_change, unknown or blocked."}, "contract": {"input_schema": {"type": "object", "additionalProperties": false, "required": ["pinned", "slots", "params"], "properties": {"pinned": {"type": "array", "minItems": 4, "maxItems": 4, "items": {"type": "object", "additionalProperties": false, "required": ["store", "key", "version", "value_sha256"], "properties": {"store": {"type": "string", "pattern": "^[a-z0-9][a-z0-9_.-]{0,63}$"}, "key": {"type": "string", "pattern": "^[A-Za-z0-9][A-Za-z0-9_.:-]{0,255}$"}, "version": {"type": "integer", "minimum": 1}, "value_sha256": {"type": "string", "pattern": "^[0-9a-f]{64}$"}}}}, "slots": {"type": "object", "additionalProperties": false, "required": ["calibration_bundle"], "properties": {"calibration_bundle": {"type": "object", "additionalProperties": false, "required": ["store", "key", "version", "value_sha256"], "properties": {"store": {"type": "string", "pattern": "^[a-z0-9][a-z0-9_.-]{0,63}$"}, "key": {"type": "string", "pattern": "^[A-Za-z0-9][A-Za-z0-9_.:-]{0,255}$"}, "version": {"type": "integer", "minimum": 1}, "value_sha256": {"type": "string", "pattern": "^[0-9a-f]{64}$"}}}}}, "params": {"type": "object", "additionalProperties": false, "required": ["output_store"], "properties": {"output_store": {"type": "string", "pattern": "^[a-z0-9][a-z0-9_.-]{0,63}$"}}}}}, "output_schema": {"type": "object", "additionalProperties": false, "required": ["outcome", "summary", "artifacts", "checks", "unknowns"], "properties": {"outcome": {"type": "string", "enum": ["rubric_judge_calibrated", "no_change", "unknown", "blocked"]}, "summary": {"type": "string", "minLength": 1, "maxLength": 1200}, "artifacts": {"type": "array", "maxItems": 1, "items": {"type": "object", "additionalProperties": false, "required": ["role", "ref"], "properties": {"role": {"type": "string", "minLength": 1, "maxLength": 64}, "ref": {"type": "object", "additionalProperties": false, "required": ["store", "key", "version", "value_sha256"], "properties": {"store": {"type": "string", "pattern": "^[a-z0-9][a-z0-9_.-]{0,63}$"}, "key": {"type": "string", "pattern": "^[A-Za-z0-9][A-Za-z0-9_.:-]{0,255}$"}, "version": {"type": "integer", "minimum": 1}, "value_sha256": {"type": "string", "pattern": "^[0-9a-f]{64}$"}}}, "value": {"type": "object"}}}}, "checks": {"type": "array", "maxItems": 64, "items": {"type": "object", "additionalProperties": false, "required": ["id", "result"], "properties": {"id": {"type": "string", "minLength": 1, "maxLength": 80}, "result": {"type": "string", "enum": ["pass", "fail", "not_run"]}, "evidence": {"type": "string", "maxLength": 2000}}}}, "unknowns": {"type": "array", "maxItems": 32, "items": {"type": "object", "additionalProperties": false, "required": ["item", "reason"], "properties": {"item": {"type": "string", "minLength": 1, "maxLength": 300}, "reason": {"type": "string", "minLength": 1, "maxLength": 1000}, "would_resolve": {"type": "string", "maxLength": 1000}}}}, "no_change_reason": {"type": "string", "minLength": 1, "maxLength": 2000}, "blocked_question": {"type": "string", "minLength": 1, "maxLength": 1000}, "proposed_verbs": {"type": "array", "maxItems": 10, "items": {"type": "object", "additionalProperties": false, "required": ["kind", "title", "outcome", "why", "priority"], "properties": {"kind": {"type": "string", "pattern": "^verb\\.[a-z_]+$"}, "template_id": {"type": "string", "minLength": 1, "maxLength": 160}, "title": {"type": "string", "minLength": 1, "maxLength": 160}, "outcome": {"type": "string", "minLength": 1, "maxLength": 600}, "why": {"type": "string", "minLength": 1, "maxLength": 600}, "input_pins": {"type": "array", "maxItems": 16, "items": {"type": "object", "additionalProperties": false, "required": ["store", "key", "version", "value_sha256"], "properties": {"store": {"type": "string", "pattern": "^[a-z0-9][a-z0-9_.-]{0,63}$"}, "key": {"type": "string", "pattern": "^[A-Za-z0-9][A-Za-z0-9_.:-]{0,255}$"}, "version": {"type": "integer", "minimum": 1}, "value_sha256": {"type": "string", "pattern": "^[0-9a-f]{64}$"}}}}, "priority": {"type": "integer", "minimum": 1, "maximum": 3}}}}, "proposed_nouns": {"type": "array", "maxItems": 10, "items": {"type": "object", "additionalProperties": false, "required": ["category", "role", "title", "why"], "properties": {"category": {"type": "string", "pattern": "^noun\\.[a-z_]+$"}, "role": {"type": "string", "minLength": 1, "maxLength": 64}, "title": {"type": "string", "minLength": 1, "maxLength": 160}, "why": {"type": "string", "minLength": 1, "maxLength": 600}}}}, "run_notes": {"type": "object", "additionalProperties": false, "properties": {"went_right": {"type": "array", "maxItems": 8, "items": {"type": "string", "minLength": 1, "maxLength": 500}}, "went_wrong": {"type": "array", "maxItems": 8, "items": {"type": "string", "minLength": 1, "maxLength": 500}}, "next_steps": {"type": "array", "maxItems": 8, "items": {"type": "string", "minLength": 1, "maxLength": 500}}, "ideas": {"type": "array", "maxItems": 8, "items": {"type": "string", "minLength": 1, "maxLength": 500}}}}}}}, "blocks": [{"kind": "error", "text": ""}], "artifact_refs": [], "usage": {"source": "harness_metadata", "input_tokens": null, "output_tokens": null, "reason": "Template only. Replace with measured counters or keep null with a reason."}}, "fill_rules": ["Copy contract.input_schema and contract.output_schema unchanged, so every run of this template shares one shape fingerprint.", "Add context_hash, environment_hash and toolchain_hash as lowercase SHA-256 digests when known; all five contract fields together allow exact matching.", "Name the exact harness version, provider and model. Deterministic or human work records the model name as none.", "Put what went wrong in an error block and in output.run_notes.went_wrong. The version-1 report has no fields for what went right, next steps or ideas, so those go in output.run_notes.", "Leave token counts null with a reason unless the harness or provider measured them."], "run_notes_field": "output.run_notes", "shape_sha256": "ee3da5c4d491d860ecd5b06f2dafe158f050384f3af9e50e1a9abfba77819e1a"}, "dimensions": {"determinism": 0.0, "evidence_grounding": 0.0, "token_efficiency": 0.0, "safety_compliance": 0.0, "verification_tier": "unverified"}, "dimension_basis": {"measured": false, "method": "unmeasured_floor", "reason": "Tessera requires these axes, so each is set to its lowest value and the tier is unverified. No quality is claimed. Replace them only with measured values and the evidence that produced them."}, "relations": [{"relation": "requires_noun", "target_id": "seedpack.self-rewarding-and-iterative-dpo-flywheel/self-rewarding-and-iterative-dpo-flywheel.v0.1.0.spec.iterative-rpo-and-flywheel-grid", "target_category": "noun.benchmark_suite", "note": ""}, {"relation": "requires_noun", "target_id": "seedpack.self-rewarding-and-iterative-dpo-flywheel/self-rewarding-and-iterative-dpo-flywheel.v0.1.0.rules.rpo-objective-and-calibration-gates", "target_category": "noun.reflex_procedure", "note": ""}, {"relation": "requires_noun", "target_id": "seedpack.self-rewarding-and-iterative-dpo-flywheel/self-rewarding-and-iterative-dpo-flywheel.v0.1.0.guardrail.block-uncalibrated-synthetic-rubric-pairs", "target_category": "noun.skill_pack", "note": ""}, {"relation": "requires_noun", "target_id": "seedpack.self-rewarding-and-iterative-dpo-flywheel/self-rewarding-and-iterative-dpo-flywheel.v0.1.0.tool.iterative-rpo-flywheel", "target_category": "noun.code_node", "note": ""}, {"relation": "produces_noun", "target_id": "seedpack.self-rewarding-and-iterative-dpo-flywheel/self-rewarding-and-iterative-dpo-flywheel.v0.1.0.template.iterative-rpo-flywheel-receipt", "target_category": "noun.scene_recipe", "note": "output calibration_receipt"}]}},
{"queue": "conformance-seed-packs", "kind": "verb.generate_seed", "effect_class": "write-internal", "priority": 50, "idempotency_key": "seedpack:self-rewarding-and-iterative-dpo-flywheel:curate-iterative-rpo-preference-pairs:f8db7286eb8c0b39898edb7f95d17ef25515c38110bc2268d6384d1151c36afb", "depends_on": [], "producer_surface": "surface:tessera", "visibility_scope": "surface_only", "allowed_surfaces": [], "target_cohorts": ["all"], "tags": ["category:verb.generate_seed", "pack:self-rewarding-and-iterative-dpo-flywheel", "template:curate-iterative-rpo-preference-pairs", "claimant:agent", "status:quarantined"], "brief": {"category_id": "verb.generate_seed", "contract": "ta-amps.seed-verb-template.v1", "pack": "self-rewarding-and-iterative-dpo-flywheel", "pack_version": "0.1.0", "template_id": "self-rewarding-and-iterative-dpo-flywheel.curate-iterative-rpo-preference-pairs", "title": "Curate preference pairs with Iterative RPO NLL regularization, length bounds and 3-key deduplication", "plain_summary": "Score (prompt, chosen_verified, rejected_tombstone) pairs with DPO logistic margin plus alpha * NLL_chosen regularization, drop chosen log-probability collapse and length-biased pairs, and deduplicate across three structural keys.", "outcome": "Curated Iterative RPO preference dataset summary and rejection reason breakdown.", "claimant": "agent", "inputs": {"pinned": [{"store": "seedpack.self-rewarding-and-iterative-dpo-flywheel", "key": "self-rewarding-and-iterative-dpo-flywheel.v0.1.0.spec.iterative-rpo-and-flywheel-grid", "version": 1, "value_sha256": "7718341ff5d936823077ad1990255647603f58735bc00d10503a9c4606b14460"}, {"store": "seedpack.self-rewarding-and-iterative-dpo-flywheel", "key": "self-rewarding-and-iterative-dpo-flywheel.v0.1.0.rules.rpo-objective-and-calibration-gates", "version": 1, "value_sha256": "22235e9bb2d7131cafbcd0b9ecb917cb86cb9104d0d3386c359ff45539c90286"}, {"store": "seedpack.self-rewarding-and-iterative-dpo-flywheel", "key": "self-rewarding-and-iterative-dpo-flywheel.v0.1.0.tool.iterative-rpo-flywheel", "version": 1, "value_sha256": "4be9841095679303522ff8e2f28440424fe39b339a5fdf7461a5beb724042130"}], "slots": {"candidate_pairs_bundle": null}, "params": {"output_store": "seedpack.self-rewarding-and-iterative-dpo-flywheel.outputs"}}, "slot_specs": {"candidate_pairs_bundle": {"category": "noun.eval_run", "required": true, "many": false, "description": "Pin of verified and calibrated preference pair candidates with policy and reference log-probabilities."}}, "param_specs": {"output_store": {"description": "Store for the produced nouns. The default is a placeholder; use the store your owner assigned.", "schema": {"type": "string", "pattern": "^[a-z0-9][a-z0-9_.-]{0,63}$"}, "required": true, "default": "seedpack.self-rewarding-and-iterative-dpo-flywheel.outputs"}}, "input_schema": {"type": "object", "additionalProperties": false, "required": ["pinned", "slots", "params"], "properties": {"pinned": {"type": "array", "minItems": 3, "maxItems": 3, "items": {"type": "object", "additionalProperties": false, "required": ["store", "key", "version", "value_sha256"], "properties": {"store": {"type": "string", "pattern": "^[a-z0-9][a-z0-9_.-]{0,63}$"}, "key": {"type": "string", "pattern": "^[A-Za-z0-9][A-Za-z0-9_.:-]{0,255}$"}, "version": {"type": "integer", "minimum": 1}, "value_sha256": {"type": "string", "pattern": "^[0-9a-f]{64}$"}}}}, "slots": {"type": "object", "additionalProperties": false, "required": ["candidate_pairs_bundle"], "properties": {"candidate_pairs_bundle": {"type": "object", "additionalProperties": false, "required": ["store", "key", "version", "value_sha256"], "properties": {"store": {"type": "string", "pattern": "^[a-z0-9][a-z0-9_.-]{0,63}$"}, "key": {"type": "string", "pattern": "^[A-Za-z0-9][A-Za-z0-9_.:-]{0,255}$"}, "version": {"type": "integer", "minimum": 1}, "value_sha256": {"type": "string", "pattern": "^[0-9a-f]{64}$"}}}}}, "params": {"type": "object", "additionalProperties": false, "required": ["output_store"], "properties": {"output_store": {"type": "string", "pattern": "^[a-z0-9][a-z0-9_.-]{0,63}$"}}}}}, "output_contract": {"visibility_scope": "surface_only", "allowed_surfaces": [], "write_mode": "create_only", "store_param": "output_store", "key_rule": "self-rewarding-and-iterative-dpo-flywheel.<role>.<task_id>[.<n>]", "expected_version": 0, "produces": [{"name": "curation_receipt", "template": {"store": "seedpack.self-rewarding-and-iterative-dpo-flywheel", "key": "self-rewarding-and-iterative-dpo-flywheel.v0.1.0.template.iterative-rpo-flywheel-receipt", "version": 1, "value_sha256": "a1f71af1350671ca8cb681c98d855404a3bdcf5bb5cec750dbabb885c621d280"}, "max": 1}], "delivery": "Write produced nouns surface_only under your own authenticated surface, which is the requester's assigned surface, so the requester and that surface's reviewers can read them. Never use public_commons or allowlist other surfaces. Return every produced noun's pin in output.artifacts; add its value when the requester asks for inline delivery."}, "output_schema": {"type": "object", "additionalProperties": false, "required": ["outcome", "summary", "artifacts", "checks", "unknowns"], "properties": {"outcome": {"type": "string", "enum": ["rpo_pairs_curated", "no_change", "unknown", "blocked"]}, "summary": {"type": "string", "minLength": 1, "maxLength": 1200}, "artifacts": {"type": "array", "maxItems": 1, "items": {"type": "object", "additionalProperties": false, "required": ["role", "ref"], "properties": {"role": {"type": "string", "minLength": 1, "maxLength": 64}, "ref": {"type": "object", "additionalProperties": false, "required": ["store", "key", "version", "value_sha256"], "properties": {"store": {"type": "string", "pattern": "^[a-z0-9][a-z0-9_.-]{0,63}$"}, "key": {"type": "string", "pattern": "^[A-Za-z0-9][A-Za-z0-9_.:-]{0,255}$"}, "version": {"type": "integer", "minimum": 1}, "value_sha256": {"type": "string", "pattern": "^[0-9a-f]{64}$"}}}, "value": {"type": "object"}}}}, "checks": {"type": "array", "maxItems": 64, "items": {"type": "object", "additionalProperties": false, "required": ["id", "result"], "properties": {"id": {"type": "string", "minLength": 1, "maxLength": 80}, "result": {"type": "string", "enum": ["pass", "fail", "not_run"]}, "evidence": {"type": "string", "maxLength": 2000}}}}, "unknowns": {"type": "array", "maxItems": 32, "items": {"type": "object", "additionalProperties": false, "required": ["item", "reason"], "properties": {"item": {"type": "string", "minLength": 1, "maxLength": 300}, "reason": {"type": "string", "minLength": 1, "maxLength": 1000}, "would_resolve": {"type": "string", "maxLength": 1000}}}}, "no_change_reason": {"type": "string", "minLength": 1, "maxLength": 2000}, "blocked_question": {"type": "string", "minLength": 1, "maxLength": 1000}, "proposed_verbs": {"type": "array", "maxItems": 10, "items": {"type": "object", "additionalProperties": false, "required": ["kind", "title", "outcome", "why", "priority"], "properties": {"kind": {"type": "string", "pattern": "^verb\\.[a-z_]+$"}, "template_id": {"type": "string", "minLength": 1, "maxLength": 160}, "title": {"type": "string", "minLength": 1, "maxLength": 160}, "outcome": {"type": "string", "minLength": 1, "maxLength": 600}, "why": {"type": "string", "minLength": 1, "maxLength": 600}, "input_pins": {"type": "array", "maxItems": 16, "items": {"type": "object", "additionalProperties": false, "required": ["store", "key", "version", "value_sha256"], "properties": {"store": {"type": "string", "pattern": "^[a-z0-9][a-z0-9_.-]{0,63}$"}, "key": {"type": "string", "pattern": "^[A-Za-z0-9][A-Za-z0-9_.:-]{0,255}$"}, "version": {"type": "integer", "minimum": 1}, "value_sha256": {"type": "string", "pattern": "^[0-9a-f]{64}$"}}}}, "priority": {"type": "integer", "minimum": 1, "maximum": 3}}}}, "proposed_nouns": {"type": "array", "maxItems": 10, "items": {"type": "object", "additionalProperties": false, "required": ["category", "role", "title", "why"], "properties": {"category": {"type": "string", "pattern": "^noun\\.[a-z_]+$"}, "role": {"type": "string", "minLength": 1, "maxLength": 64}, "title": {"type": "string", "minLength": 1, "maxLength": 160}, "why": {"type": "string", "minLength": 1, "maxLength": 600}}}}, "run_notes": {"type": "object", "additionalProperties": false, "properties": {"went_right": {"type": "array", "maxItems": 8, "items": {"type": "string", "minLength": 1, "maxLength": 500}}, "went_wrong": {"type": "array", "maxItems": 8, "items": {"type": "string", "minLength": 1, "maxLength": 500}}, "next_steps": {"type": "array", "maxItems": 8, "items": {"type": "string", "minLength": 1, "maxLength": 500}}, "ideas": {"type": "array", "maxItems": 8, "items": {"type": "string", "minLength": 1, "maxLength": 500}}}}}}, "permitted_outcomes": ["rpo_pairs_curated", "no_change", "unknown", "blocked"], "outcome_meanings": {"rpo_pairs_curated": "The declared nouns were produced and every deterministic acceptance check passed.", "no_change": "The evidence does not justify producing or changing anything. Say why.", "unknown": "The question could not be answered with the permitted inputs. List what is missing and what would resolve it.", "blocked": "A required input, permission or capability is missing. Ask one precise question."}, "outcome_requirements": {"rpo_pairs_curated": ["artifacts"], "no_change": ["no_change_reason"], "unknown": ["unknowns"], "blocked": ["blocked_question"]}, "acceptance_checks": [{"id": "rpo-collapse-length-and-dedup-guards", "check": "All admitted pairs satisfy length-ratio bounds, non-collapsed chosen logp drift, positive DPO margin and 3-key uniqueness.", "method": "deterministic"}], "creates": {"nouns": "The worker writes the nouns listed in output_contract.produces and returns their pins.", "verbs": "The worker proposes follow-up verbs in output.proposed_verbs, such as a breakdown into smaller verbs or a research follow-up, with the pins they need. Proposals are never enqueued by the worker; an authorized requester enqueues the ones worth doing."}, "guidance": {"required": ["Compute DPO implicit reward margins, per-token NLL_chosen regularizers and chosen log-probability drift for each candidate pair.", "Filter unverified, length-biased, logp-collapsed and 3-key structurally duplicated pairs and rank survivors by Iterative RPO loss.", "Return follow-up work the results justify as proposed_verbs and nouns worth creating later as proposed_nouns. Never enqueue them yourself; an authorized requester decides which to run."], "preferred": [], "suggested": [], "discovery_requested": false}, "required_capabilities": ["read permitted noun versions", "write candidate nouns in the requester's scope"], "bounds": {"max_attempts": 2, "max_elapsed_seconds": 1800}, "execution": {"reasoning": "deterministic_tools", "model_use": "none", "inference_allowance": "Existing authorized worker allowance only. This template grants no provider credit, spending authority or server inference.", "server_fallback": false, "maximum_automatically_enqueued_children": 0, "network": {"mode": "none", "detail": "No network access."}, "external_effects": "none"}, "review": {"required": "owning_policy", "self_approval_permitted": false, "publication_permitted": false, "admission_permitted": false}, "instantiation": {"status": "template", "fill_before_enqueue": ["candidate_pairs_bundle"], "rule": "Replace each slot with exact NounRef pins, set the params, then recompute the idempotency key with scripts/build_packs.py --instantiate. Enqueue only with your own grants in a queue your owner assigned."}, "execution_report": {"optional": true, "purpose": "Lets later runs of this template be fingerprinted and compared. It is worker reported, not proof.", "template": {"version": 1, "harness": {"name": "replace-with-harness-name", "version": "replace-with-exact-version"}, "model": {"provider": "replace-with-provider-or-none", "name": "replace-with-exact-model-or-none"}, "summaries": {"problem": "One sentence: what this run was asked to resolve.", "input": "One sentence: which pinned inputs were read.", "solution": "One sentence: what was produced or why the outcome was no_change, unknown or blocked."}, "contract": {"input_schema": {"type": "object", "additionalProperties": false, "required": ["pinned", "slots", "params"], "properties": {"pinned": {"type": "array", "minItems": 3, "maxItems": 3, "items": {"type": "object", "additionalProperties": false, "required": ["store", "key", "version", "value_sha256"], "properties": {"store": {"type": "string", "pattern": "^[a-z0-9][a-z0-9_.-]{0,63}$"}, "key": {"type": "string", "pattern": "^[A-Za-z0-9][A-Za-z0-9_.:-]{0,255}$"}, "version": {"type": "integer", "minimum": 1}, "value_sha256": {"type": "string", "pattern": "^[0-9a-f]{64}$"}}}}, "slots": {"type": "object", "additionalProperties": false, "required": ["candidate_pairs_bundle"], "properties": {"candidate_pairs_bundle": {"type": "object", "additionalProperties": false, "required": ["store", "key", "version", "value_sha256"], "properties": {"store": {"type": "string", "pattern": "^[a-z0-9][a-z0-9_.-]{0,63}$"}, "key": {"type": "string", "pattern": "^[A-Za-z0-9][A-Za-z0-9_.:-]{0,255}$"}, "version": {"type": "integer", "minimum": 1}, "value_sha256": {"type": "string", "pattern": "^[0-9a-f]{64}$"}}}}}, "params": {"type": "object", "additionalProperties": false, "required": ["output_store"], "properties": {"output_store": {"type": "string", "pattern": "^[a-z0-9][a-z0-9_.-]{0,63}$"}}}}}, "output_schema": {"type": "object", "additionalProperties": false, "required": ["outcome", "summary", "artifacts", "checks", "unknowns"], "properties": {"outcome": {"type": "string", "enum": ["rpo_pairs_curated", "no_change", "unknown", "blocked"]}, "summary": {"type": "string", "minLength": 1, "maxLength": 1200}, "artifacts": {"type": "array", "maxItems": 1, "items": {"type": "object", "additionalProperties": false, "required": ["role", "ref"], "properties": {"role": {"type": "string", "minLength": 1, "maxLength": 64}, "ref": {"type": "object", "additionalProperties": false, "required": ["store", "key", "version", "value_sha256"], "properties": {"store": {"type": "string", "pattern": "^[a-z0-9][a-z0-9_.-]{0,63}$"}, "key": {"type": "string", "pattern": "^[A-Za-z0-9][A-Za-z0-9_.:-]{0,255}$"}, "version": {"type": "integer", "minimum": 1}, "value_sha256": {"type": "string", "pattern": "^[0-9a-f]{64}$"}}}, "value": {"type": "object"}}}}, "checks": {"type": "array", "maxItems": 64, "items": {"type": "object", "additionalProperties": false, "required": ["id", "result"], "properties": {"id": {"type": "string", "minLength": 1, "maxLength": 80}, "result": {"type": "string", "enum": ["pass", "fail", "not_run"]}, "evidence": {"type": "string", "maxLength": 2000}}}}, "unknowns": {"type": "array", "maxItems": 32, "items": {"type": "object", "additionalProperties": false, "required": ["item", "reason"], "properties": {"item": {"type": "string", "minLength": 1, "maxLength": 300}, "reason": {"type": "string", "minLength": 1, "maxLength": 1000}, "would_resolve": {"type": "string", "maxLength": 1000}}}}, "no_change_reason": {"type": "string", "minLength": 1, "maxLength": 2000}, "blocked_question": {"type": "string", "minLength": 1, "maxLength": 1000}, "proposed_verbs": {"type": "array", "maxItems": 10, "items": {"type": "object", "additionalProperties": false, "required": ["kind", "title", "outcome", "why", "priority"], "properties": {"kind": {"type": "string", "pattern": "^verb\\.[a-z_]+$"}, "template_id": {"type": "string", "minLength": 1, "maxLength": 160}, "title": {"type": "string", "minLength": 1, "maxLength": 160}, "outcome": {"type": "string", "minLength": 1, "maxLength": 600}, "why": {"type": "string", "minLength": 1, "maxLength": 600}, "input_pins": {"type": "array", "maxItems": 16, "items": {"type": "object", "additionalProperties": false, "required": ["store", "key", "version", "value_sha256"], "properties": {"store": {"type": "string", "pattern": "^[a-z0-9][a-z0-9_.-]{0,63}$"}, "key": {"type": "string", "pattern": "^[A-Za-z0-9][A-Za-z0-9_.:-]{0,255}$"}, "version": {"type": "integer", "minimum": 1}, "value_sha256": {"type": "string", "pattern": "^[0-9a-f]{64}$"}}}}, "priority": {"type": "integer", "minimum": 1, "maximum": 3}}}}, "proposed_nouns": {"type": "array", "maxItems": 10, "items": {"type": "object", "additionalProperties": false, "required": ["category", "role", "title", "why"], "properties": {"category": {"type": "string", "pattern": "^noun\\.[a-z_]+$"}, "role": {"type": "string", "minLength": 1, "maxLength": 64}, "title": {"type": "string", "minLength": 1, "maxLength": 160}, "why": {"type": "string", "minLength": 1, "maxLength": 600}}}}, "run_notes": {"type": "object", "additionalProperties": false, "properties": {"went_right": {"type": "array", "maxItems": 8, "items": {"type": "string", "minLength": 1, "maxLength": 500}}, "went_wrong": {"type": "array", "maxItems": 8, "items": {"type": "string", "minLength": 1, "maxLength": 500}}, "next_steps": {"type": "array", "maxItems": 8, "items": {"type": "string", "minLength": 1, "maxLength": 500}}, "ideas": {"type": "array", "maxItems": 8, "items": {"type": "string", "minLength": 1, "maxLength": 500}}}}}}}, "blocks": [{"kind": "error", "text": ""}], "artifact_refs": [], "usage": {"source": "harness_metadata", "input_tokens": null, "output_tokens": null, "reason": "Template only. Replace with measured counters or keep null with a reason."}}, "fill_rules": ["Copy contract.input_schema and contract.output_schema unchanged, so every run of this template shares one shape fingerprint.", "Add context_hash, environment_hash and toolchain_hash as lowercase SHA-256 digests when known; all five contract fields together allow exact matching.", "Name the exact harness version, provider and model. Deterministic or human work records the model name as none.", "Put what went wrong in an error block and in output.run_notes.went_wrong. The version-1 report has no fields for what went right, next steps or ideas, so those go in output.run_notes.", "Leave token counts null with a reason unless the harness or provider measured them."], "run_notes_field": "output.run_notes", "shape_sha256": "f02bf4d8a98469e9e5d71cbb649619800c8b355d7f23a53ff4f77939a71fc3f0"}, "dimensions": {"determinism": 0.0, "evidence_grounding": 0.0, "token_efficiency": 0.0, "safety_compliance": 0.0, "verification_tier": "unverified"}, "dimension_basis": {"measured": false, "method": "unmeasured_floor", "reason": "Tessera requires these axes, so each is set to its lowest value and the tier is unverified. No quality is claimed. Replace them only with measured values and the evidence that produced them."}, "relations": [{"relation": "requires_noun", "target_id": "seedpack.self-rewarding-and-iterative-dpo-flywheel/self-rewarding-and-iterative-dpo-flywheel.v0.1.0.spec.iterative-rpo-and-flywheel-grid", "target_category": "noun.benchmark_suite", "note": ""}, {"relation": "requires_noun", "target_id": "seedpack.self-rewarding-and-iterative-dpo-flywheel/self-rewarding-and-iterative-dpo-flywheel.v0.1.0.rules.rpo-objective-and-calibration-gates", "target_category": "noun.reflex_procedure", "note": ""}, {"relation": "requires_noun", "target_id": "seedpack.self-rewarding-and-iterative-dpo-flywheel/self-rewarding-and-iterative-dpo-flywheel.v0.1.0.tool.iterative-rpo-flywheel", "target_category": "noun.code_node", "note": ""}, {"relation": "produces_noun", "target_id": "seedpack.self-rewarding-and-iterative-dpo-flywheel/self-rewarding-and-iterative-dpo-flywheel.v0.1.0.template.iterative-rpo-flywheel-receipt", "target_category": "noun.scene_recipe", "note": "output curation_receipt"}], "follows_templates": ["self-rewarding-and-iterative-dpo-flywheel.calibrate-rubric-judge-against-verifiers"]}},
{"queue": "conformance-seed-packs", "kind": "verb.verify_gate", "effect_class": "write-internal", "priority": 50, "idempotency_key": "seedpack:self-rewarding-and-iterative-dpo-flywheel:verify-preregistered-flywheel-promotion:ff1070d666b8de31a19587deb2de5a51640a53414eab78f8f21e0ad87a7cefcd", "depends_on": [], "producer_surface": "surface:tessera", "visibility_scope": "surface_only", "allowed_surfaces": [], "target_cohorts": ["all"], "tags": ["category:verb.verify_gate", "pack:self-rewarding-and-iterative-dpo-flywheel", "template:verify-preregistered-flywheel-promotion", "claimant:agent", "status:quarantined"], "brief": {"category_id": "verb.verify_gate", "contract": "ta-amps.seed-verb-template.v1", "pack": "self-rewarding-and-iterative-dpo-flywheel", "pack_version": "0.1.0", "template_id": "self-rewarding-and-iterative-dpo-flywheel.verify-preregistered-flywheel-promotion", "title": "Independently verify holdout SHA-256 locks, regression caps and AST magic-literal gates before flywheel promotion", "plain_summary": "Verify candidate generation rounds against pre-registered holdout SHA-256 digests, sanity retention floors, regression caps, distinct reviewer identity and ZeroHardcodedValuesAstGate conformance.", "outcome": "Independent deterministic flywheel promotion receipt.", "claimant": "agent", "inputs": {"pinned": [{"store": "seedpack.self-rewarding-and-iterative-dpo-flywheel", "key": "self-rewarding-and-iterative-dpo-flywheel.v0.1.0.spec.iterative-rpo-and-flywheel-grid", "version": 1, "value_sha256": "7718341ff5d936823077ad1990255647603f58735bc00d10503a9c4606b14460"}, {"store": "seedpack.self-rewarding-and-iterative-dpo-flywheel", "key": "self-rewarding-and-iterative-dpo-flywheel.v0.1.0.guardrail.block-uncalibrated-synthetic-rubric-pairs", "version": 1, "value_sha256": "7cb46a0edcc6becf59dc750a792ab8e46ba519b613cf9ee2015c7b54e24df548"}, {"store": "seedpack.self-rewarding-and-iterative-dpo-flywheel", "key": "self-rewarding-and-iterative-dpo-flywheel.v0.1.0.guardrail.preregistered-holdout-lock-and-zero-magic-literals", "version": 1, "value_sha256": "9a8a38847ea3240ef297623357b91de674113fd803e2fc9c59165da601011cc7"}, {"store": "seedpack.self-rewarding-and-iterative-dpo-flywheel", "key": "self-rewarding-and-iterative-dpo-flywheel.v0.1.0.tool.iterative-rpo-flywheel", "version": 1, "value_sha256": "4be9841095679303522ff8e2f28440424fe39b339a5fdf7461a5beb724042130"}, {"store": "seedpack.self-rewarding-and-iterative-dpo-flywheel", "key": "self-rewarding-and-iterative-dpo-flywheel.v0.1.0.template.iterative-rpo-flywheel-receipt", "version": 1, "value_sha256": "a1f71af1350671ca8cb681c98d855404a3bdcf5bb5cec750dbabb885c621d280"}], "slots": {"flywheel_eval_bundle": null}, "params": {"output_store": "seedpack.self-rewarding-and-iterative-dpo-flywheel.outputs"}}, "slot_specs": {"flywheel_eval_bundle": {"category": "noun.eval_run", "required": true, "many": false, "description": "Pin of holdout suite, baseline/candidate outcomes, sanity canaries and curation bundle."}}, "param_specs": {"output_store": {"description": "Store for the produced nouns. The default is a placeholder; use the store your owner assigned.", "schema": {"type": "string", "pattern": "^[a-z0-9][a-z0-9_.-]{0,63}$"}, "required": true, "default": "seedpack.self-rewarding-and-iterative-dpo-flywheel.outputs"}}, "input_schema": {"type": "object", "additionalProperties": false, "required": ["pinned", "slots", "params"], "properties": {"pinned": {"type": "array", "minItems": 5, "maxItems": 5, "items": {"type": "object", "additionalProperties": false, "required": ["store", "key", "version", "value_sha256"], "properties": {"store": {"type": "string", "pattern": "^[a-z0-9][a-z0-9_.-]{0,63}$"}, "key": {"type": "string", "pattern": "^[A-Za-z0-9][A-Za-z0-9_.:-]{0,255}$"}, "version": {"type": "integer", "minimum": 1}, "value_sha256": {"type": "string", "pattern": "^[0-9a-f]{64}$"}}}}, "slots": {"type": "object", "additionalProperties": false, "required": ["flywheel_eval_bundle"], "properties": {"flywheel_eval_bundle": {"type": "object", "additionalProperties": false, "required": ["store", "key", "version", "value_sha256"], "properties": {"store": {"type": "string", "pattern": "^[a-z0-9][a-z0-9_.-]{0,63}$"}, "key": {"type": "string", "pattern": "^[A-Za-z0-9][A-Za-z0-9_.:-]{0,255}$"}, "version": {"type": "integer", "minimum": 1}, "value_sha256": {"type": "string", "pattern": "^[0-9a-f]{64}$"}}}}}, "params": {"type": "object", "additionalProperties": false, "required": ["output_store"], "properties": {"output_store": {"type": "string", "pattern": "^[a-z0-9][a-z0-9_.-]{0,63}$"}}}}}, "output_contract": {"visibility_scope": "surface_only", "allowed_surfaces": [], "write_mode": "create_only", "store_param": "output_store", "key_rule": "self-rewarding-and-iterative-dpo-flywheel.<role>.<task_id>[.<n>]", "expected_version": 0, "produces": [{"name": "verified_flywheel_receipt", "template": {"store": "seedpack.self-rewarding-and-iterative-dpo-flywheel", "key": "self-rewarding-and-iterative-dpo-flywheel.v0.1.0.template.iterative-rpo-flywheel-receipt", "version": 1, "value_sha256": "a1f71af1350671ca8cb681c98d855404a3bdcf5bb5cec750dbabb885c621d280"}, "max": 1}], "delivery": "Write produced nouns surface_only under your own authenticated surface, which is the requester's assigned surface, so the requester and that surface's reviewers can read them. Never use public_commons or allowlist other surfaces. Return every produced noun's pin in output.artifacts; add its value when the requester asks for inline delivery."}, "output_schema": {"type": "object", "additionalProperties": false, "required": ["outcome", "summary", "artifacts", "checks", "unknowns"], "properties": {"outcome": {"type": "string", "enum": ["flywheel_round_verified", "no_change", "unknown", "blocked"]}, "summary": {"type": "string", "minLength": 1, "maxLength": 1200}, "artifacts": {"type": "array", "maxItems": 1, "items": {"type": "object", "additionalProperties": false, "required": ["role", "ref"], "properties": {"role": {"type": "string", "minLength": 1, "maxLength": 64}, "ref": {"type": "object", "additionalProperties": false, "required": ["store", "key", "version", "value_sha256"], "properties": {"store": {"type": "string", "pattern": "^[a-z0-9][a-z0-9_.-]{0,63}$"}, "key": {"type": "string", "pattern": "^[A-Za-z0-9][A-Za-z0-9_.:-]{0,255}$"}, "version": {"type": "integer", "minimum": 1}, "value_sha256": {"type": "string", "pattern": "^[0-9a-f]{64}$"}}}, "value": {"type": "object"}}}}, "checks": {"type": "array", "maxItems": 64, "items": {"type": "object", "additionalProperties": false, "required": ["id", "result"], "properties": {"id": {"type": "string", "minLength": 1, "maxLength": 80}, "result": {"type": "string", "enum": ["pass", "fail", "not_run"]}, "evidence": {"type": "string", "maxLength": 2000}}}}, "unknowns": {"type": "array", "maxItems": 32, "items": {"type": "object", "additionalProperties": false, "required": ["item", "reason"], "properties": {"item": {"type": "string", "minLength": 1, "maxLength": 300}, "reason": {"type": "string", "minLength": 1, "maxLength": 1000}, "would_resolve": {"type": "string", "maxLength": 1000}}}}, "no_change_reason": {"type": "string", "minLength": 1, "maxLength": 2000}, "blocked_question": {"type": "string", "minLength": 1, "maxLength": 1000}, "proposed_verbs": {"type": "array", "maxItems": 10, "items": {"type": "object", "additionalProperties": false, "required": ["kind", "title", "outcome", "why", "priority"], "properties": {"kind": {"type": "string", "pattern": "^verb\\.[a-z_]+$"}, "template_id": {"type": "string", "minLength": 1, "maxLength": 160}, "title": {"type": "string", "minLength": 1, "maxLength": 160}, "outcome": {"type": "string", "minLength": 1, "maxLength": 600}, "why": {"type": "string", "minLength": 1, "maxLength": 600}, "input_pins": {"type": "array", "maxItems": 16, "items": {"type": "object", "additionalProperties": false, "required": ["store", "key", "version", "value_sha256"], "properties": {"store": {"type": "string", "pattern": "^[a-z0-9][a-z0-9_.-]{0,63}$"}, "key": {"type": "string", "pattern": "^[A-Za-z0-9][A-Za-z0-9_.:-]{0,255}$"}, "version": {"type": "integer", "minimum": 1}, "value_sha256": {"type": "string", "pattern": "^[0-9a-f]{64}$"}}}}, "priority": {"type": "integer", "minimum": 1, "maximum": 3}}}}, "proposed_nouns": {"type": "array", "maxItems": 10, "items": {"type": "object", "additionalProperties": false, "required": ["category", "role", "title", "why"], "properties": {"category": {"type": "string", "pattern": "^noun\\.[a-z_]+$"}, "role": {"type": "string", "minLength": 1, "maxLength": 64}, "title": {"type": "string", "minLength": 1, "maxLength": 160}, "why": {"type": "string", "minLength": 1, "maxLength": 600}}}}, "run_notes": {"type": "object", "additionalProperties": false, "properties": {"went_right": {"type": "array", "maxItems": 8, "items": {"type": "string", "minLength": 1, "maxLength": 500}}, "went_wrong": {"type": "array", "maxItems": 8, "items": {"type": "string", "minLength": 1, "maxLength": 500}}, "next_steps": {"type": "array", "maxItems": 8, "items": {"type": "string", "minLength": 1, "maxLength": 500}}, "ideas": {"type": "array", "maxItems": 8, "items": {"type": "string", "minLength": 1, "maxLength": 500}}}}}}, "permitted_outcomes": ["flywheel_round_verified", "no_change", "unknown", "blocked"], "outcome_meanings": {"flywheel_round_verified": "The declared nouns were produced and every deterministic acceptance check passed.", "no_change": "The evidence does not justify producing or changing anything. Say why.", "unknown": "The question could not be answered with the permitted inputs. List what is missing and what would resolve it.", "blocked": "A required input, permission or capability is missing. Ask one precise question."}, "outcome_requirements": {"flywheel_round_verified": ["artifacts"], "no_change": ["no_change_reason"], "unknown": ["unknowns"], "blocked": ["blocked_question"]}, "acceptance_checks": [{"id": "holdout-lock-and-ast-conformance", "check": "scan_directory_for_hardcoded_values returns an empty list, holdout SHA-256 matches and run_bundle reproduces the exact template-conforming receipt.", "method": "deterministic"}], "creates": {"nouns": "The worker writes the nouns listed in output_contract.produces and returns their pins.", "verbs": "The worker proposes follow-up verbs in output.proposed_verbs, such as a breakdown into smaller verbs or a research follow-up, with the pins they need. Proposals are never enqueued by the worker; an authorized requester enqueues the ones worth doing."}, "guidance": {"required": ["Verify canonical SHA-256 equality on the holdout suite, distinct proposer/reviewer identifiers, sanity retention and regression rate caps.", "Run scan_directory_for_hardcoded_values across iterative_rpo_flywheel and confirm zero AST magic-literal violations.", "Return follow-up work the results justify as proposed_verbs and nouns worth creating later as proposed_nouns. Never enqueue them yourself; an authorized requester decides which to run."], "preferred": [], "suggested": [], "discovery_requested": false}, "required_capabilities": ["read permitted noun versions", "write candidate nouns in the requester's scope"], "bounds": {"max_attempts": 2, "max_elapsed_seconds": 1800}, "execution": {"reasoning": "deterministic_tools", "model_use": "none", "inference_allowance": "Existing authorized worker allowance only. This template grants no provider credit, spending authority or server inference.", "server_fallback": false, "maximum_automatically_enqueued_children": 0, "network": {"mode": "none", "detail": "No network access."}, "external_effects": "none"}, "review": {"required": "distinct_reviewer", "self_approval_permitted": false, "publication_permitted": false, "admission_permitted": false}, "instantiation": {"status": "template", "fill_before_enqueue": ["flywheel_eval_bundle"], "rule": "Replace each slot with exact NounRef pins, set the params, then recompute the idempotency key with scripts/build_packs.py --instantiate. Enqueue only with your own grants in a queue your owner assigned."}, "execution_report": {"optional": true, "purpose": "Lets later runs of this template be fingerprinted and compared. It is worker reported, not proof.", "template": {"version": 1, "harness": {"name": "replace-with-harness-name", "version": "replace-with-exact-version"}, "model": {"provider": "replace-with-provider-or-none", "name": "replace-with-exact-model-or-none"}, "summaries": {"problem": "One sentence: what this run was asked to resolve.", "input": "One sentence: which pinned inputs were read.", "solution": "One sentence: what was produced or why the outcome was no_change, unknown or blocked."}, "contract": {"input_schema": {"type": "object", "additionalProperties": false, "required": ["pinned", "slots", "params"], "properties": {"pinned": {"type": "array", "minItems": 5, "maxItems": 5, "items": {"type": "object", "additionalProperties": false, "required": ["store", "key", "version", "value_sha256"], "properties": {"store": {"type": "string", "pattern": "^[a-z0-9][a-z0-9_.-]{0,63}$"}, "key": {"type": "string", "pattern": "^[A-Za-z0-9][A-Za-z0-9_.:-]{0,255}$"}, "version": {"type": "integer", "minimum": 1}, "value_sha256": {"type": "string", "pattern": "^[0-9a-f]{64}$"}}}}, "slots": {"type": "object", "additionalProperties": false, "required": ["flywheel_eval_bundle"], "properties": {"flywheel_eval_bundle": {"type": "object", "additionalProperties": false, "required": ["store", "key", "version", "value_sha256"], "properties": {"store": {"type": "string", "pattern": "^[a-z0-9][a-z0-9_.-]{0,63}$"}, "key": {"type": "string", "pattern": "^[A-Za-z0-9][A-Za-z0-9_.:-]{0,255}$"}, "version": {"type": "integer", "minimum": 1}, "value_sha256": {"type": "string", "pattern": "^[0-9a-f]{64}$"}}}}}, "params": {"type": "object", "additionalProperties": false, "required": ["output_store"], "properties": {"output_store": {"type": "string", "pattern": "^[a-z0-9][a-z0-9_.-]{0,63}$"}}}}}, "output_schema": {"type": "object", "additionalProperties": false, "required": ["outcome", "summary", "artifacts", "checks", "unknowns"], "properties": {"outcome": {"type": "string", "enum": ["flywheel_round_verified", "no_change", "unknown", "blocked"]}, "summary": {"type": "string", "minLength": 1, "maxLength": 1200}, "artifacts": {"type": "array", "maxItems": 1, "items": {"type": "object", "additionalProperties": false, "required": ["role", "ref"], "properties": {"role": {"type": "string", "minLength": 1, "maxLength": 64}, "ref": {"type": "object", "additionalProperties": false, "required": ["store", "key", "version", "value_sha256"], "properties": {"store": {"type": "string", "pattern": "^[a-z0-9][a-z0-9_.-]{0,63}$"}, "key": {"type": "string", "pattern": "^[A-Za-z0-9][A-Za-z0-9_.:-]{0,255}$"}, "version": {"type": "integer", "minimum": 1}, "value_sha256": {"type": "string", "pattern": "^[0-9a-f]{64}$"}}}, "value": {"type": "object"}}}}, "checks": {"type": "array", "maxItems": 64, "items": {"type": "object", "additionalProperties": false, "required": ["id", "result"], "properties": {"id": {"type": "string", "minLength": 1, "maxLength": 80}, "result": {"type": "string", "enum": ["pass", "fail", "not_run"]}, "evidence": {"type": "string", "maxLength": 2000}}}}, "unknowns": {"type": "array", "maxItems": 32, "items": {"type": "object", "additionalProperties": false, "required": ["item", "reason"], "properties": {"item": {"type": "string", "minLength": 1, "maxLength": 300}, "reason": {"type": "string", "minLength": 1, "maxLength": 1000}, "would_resolve": {"type": "string", "maxLength": 1000}}}}, "no_change_reason": {"type": "string", "minLength": 1, "maxLength": 2000}, "blocked_question": {"type": "string", "minLength": 1, "maxLength": 1000}, "proposed_verbs": {"type": "array", "maxItems": 10, "items": {"type": "object", "additionalProperties": false, "required": ["kind", "title", "outcome", "why", "priority"], "properties": {"kind": {"type": "string", "pattern": "^verb\\.[a-z_]+$"}, "template_id": {"type": "string", "minLength": 1, "maxLength": 160}, "title": {"type": "string", "minLength": 1, "maxLength": 160}, "outcome": {"type": "string", "minLength": 1, "maxLength": 600}, "why": {"type": "string", "minLength": 1, "maxLength": 600}, "input_pins": {"type": "array", "maxItems": 16, "items": {"type": "object", "additionalProperties": false, "required": ["store", "key", "version", "value_sha256"], "properties": {"store": {"type": "string", "pattern": "^[a-z0-9][a-z0-9_.-]{0,63}$"}, "key": {"type": "string", "pattern": "^[A-Za-z0-9][A-Za-z0-9_.:-]{0,255}$"}, "version": {"type": "integer", "minimum": 1}, "value_sha256": {"type": "string", "pattern": "^[0-9a-f]{64}$"}}}}, "priority": {"type": "integer", "minimum": 1, "maximum": 3}}}}, "proposed_nouns": {"type": "array", "maxItems": 10, "items": {"type": "object", "additionalProperties": false, "required": ["category", "role", "title", "why"], "properties": {"category": {"type": "string", "pattern": "^noun\\.[a-z_]+$"}, "role": {"type": "string", "minLength": 1, "maxLength": 64}, "title": {"type": "string", "minLength": 1, "maxLength": 160}, "why": {"type": "string", "minLength": 1, "maxLength": 600}}}}, "run_notes": {"type": "object", "additionalProperties": false, "properties": {"went_right": {"type": "array", "maxItems": 8, "items": {"type": "string", "minLength": 1, "maxLength": 500}}, "went_wrong": {"type": "array", "maxItems": 8, "items": {"type": "string", "minLength": 1, "maxLength": 500}}, "next_steps": {"type": "array", "maxItems": 8, "items": {"type": "string", "minLength": 1, "maxLength": 500}}, "ideas": {"type": "array", "maxItems": 8, "items": {"type": "string", "minLength": 1, "maxLength": 500}}}}}}}, "blocks": [{"kind": "error", "text": ""}], "artifact_refs": [], "usage": {"source": "harness_metadata", "input_tokens": null, "output_tokens": null, "reason": "Template only. Replace with measured counters or keep null with a reason."}}, "fill_rules": ["Copy contract.input_schema and contract.output_schema unchanged, so every run of this template shares one shape fingerprint.", "Add context_hash, environment_hash and toolchain_hash as lowercase SHA-256 digests when known; all five contract fields together allow exact matching.", "Name the exact harness version, provider and model. Deterministic or human work records the model name as none.", "Put what went wrong in an error block and in output.run_notes.went_wrong. The version-1 report has no fields for what went right, next steps or ideas, so those go in output.run_notes.", "Leave token counts null with a reason unless the harness or provider measured them."], "run_notes_field": "output.run_notes", "shape_sha256": "239eada750ba0f5f8507722c2b97cd72c9e7eeb4893e41c18fe6d2c46ef6d3ee"}, "dimensions": {"determinism": 0.0, "evidence_grounding": 0.0, "token_efficiency": 0.0, "safety_compliance": 0.0, "verification_tier": "unverified"}, "dimension_basis": {"measured": false, "method": "unmeasured_floor", "reason": "Tessera requires these axes, so each is set to its lowest value and the tier is unverified. No quality is claimed. Replace them only with measured values and the evidence that produced them."}, "relations": [{"relation": "requires_noun", "target_id": "seedpack.self-rewarding-and-iterative-dpo-flywheel/self-rewarding-and-iterative-dpo-flywheel.v0.1.0.spec.iterative-rpo-and-flywheel-grid", "target_category": "noun.benchmark_suite", "note": ""}, {"relation": "requires_noun", "target_id": "seedpack.self-rewarding-and-iterative-dpo-flywheel/self-rewarding-and-iterative-dpo-flywheel.v0.1.0.guardrail.block-uncalibrated-synthetic-rubric-pairs", "target_category": "noun.skill_pack", "note": ""}, {"relation": "requires_noun", "target_id": "seedpack.self-rewarding-and-iterative-dpo-flywheel/self-rewarding-and-iterative-dpo-flywheel.v0.1.0.guardrail.preregistered-holdout-lock-and-zero-magic-literals", "target_category": "noun.skill_pack", "note": ""}, {"relation": "requires_noun", "target_id": "seedpack.self-rewarding-and-iterative-dpo-flywheel/self-rewarding-and-iterative-dpo-flywheel.v0.1.0.tool.iterative-rpo-flywheel", "target_category": "noun.code_node", "note": ""}, {"relation": "requires_noun", "target_id": "seedpack.self-rewarding-and-iterative-dpo-flywheel/self-rewarding-and-iterative-dpo-flywheel.v0.1.0.template.iterative-rpo-flywheel-receipt", "target_category": "noun.scene_recipe", "note": ""}, {"relation": "produces_noun", "target_id": "seedpack.self-rewarding-and-iterative-dpo-flywheel/self-rewarding-and-iterative-dpo-flywheel.v0.1.0.template.iterative-rpo-flywheel-receipt", "target_category": "noun.scene_recipe", "note": "output verified_flywheel_receipt"}], "follows_templates": ["self-rewarding-and-iterative-dpo-flywheel.curate-iterative-rpo-preference-pairs"]}}
]}