[{"_1":2,"_3":-5,"_4":-5},"loaderData",{"_5":6,"_7":8,"_9":10},"actionData","errors","root",{"_160":161,"_162":163},"routes/_app.x.$handle.$name",{"_100":99,"_32":45,"_109":90,"_27":110,"_13":14,"_111":-5,"_11":112,"_113":114,"_115":116,"_117":118,"_119":120,"_20":-5,"_121":-5,"_122":-5,"_123":25,"_124":125,"_126":25,"_127":-5,"_128":-5,"_129":130},"routes/_app.x.$handle.$name.branch.$branch.version.$version.diff",{"_11":12,"_13":14,"_15":16,"_17":-5,"_18":-5},"project",{"_43":44,"_32":45,"_38":46,"_47":-5,"_48":25,"_49":50,"_51":50,"_52":53,"_54":-5,"_18":55,"_56":57,"_58":59,"_60":-5,"_61":-5,"_62":63,"_64":65,"_66":67,"_68":69,"_70":71,"_72":-5,"_73":74,"_75":74,"_76":77,"_78":77,"_79":77,"_80":74,"_81":74,"_82":83,"_84":85,"_86":87,"_88":25},"version",9,"changes",[19],"aiChatInfo","userId",{"_20":21,"_22":23,"_24":25,"_26":25,"_27":28,"_29":28},"file",{"_30":31,"_32":28,"_33":34,"_35":37,"_38":39,"_40":14,"_41":42,"_24":25},"previous",{"_30":31,"_32":28,"_33":34,"_35":36},"deleted",false,"deletedHere","path","candidate_v1.ts","latestPath","fileId","80e7febe-abbd-11f1-92c3-1607ee4eb77e","name","fileType","script","content","import { ValTownOpenAI } from \"https://esm.town/v/std/openai/main.ts\";\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nimport { blob as globalBlob } from \"https://esm.town/v/std/blob/global.ts\";\n\nconst H_TOKEN=\"https://mosadek--f0167112abb811f19b3f1607ee4eb77e.web.val.run\";\nconst H_ROUTE=\"https://mosadek--b1a23924abb911f1b6d51607ee4eb77e.web.val.run\";\nconst openai=new ValTownOpenAI({});\nconst candidateId=\"VAL_NATIVE_GENERIC_JIT_CONTROLLER_V1R1\";\nconst runId=crypto.randomUUID();\nconst transcript:any[]=[];\n\nasync function req(base:string,path:string,init?:RequestInit){\n const r=await fetch(base+path,init);\n const text=await r.text();\n let json:any=null; try{json=JSON.parse(text)}catch{}\n transcript.push({kind:\"HTTP\",base,path,status:r.status,json});\n return{status:r.status,json,text};\n}\nasync function post(base:string,path:string,body:any){\n return req(base,path,{method:\"POST\",headers:{\"content-type\":\"application/json\"},body:JSON.stringify(body)});\n}\nfunction assert(x:any,msg:string){if(!x)throw new Error(msg)}\nfunction compactManifest(x:any){return{\n id:x?.id??x?.skill_id??null,\n name:x?.name??null,\n description:x?.description??null,\n input_kind:x?.input_kind??null,\n output_kind:x?.output_kind??null,\n input_schema:x?.input_schema??x?.inputSchema??x?.input??null,\n output_schema:x?.output_schema??x?.outputSchema??x?.output??null,\n args:x?.args??null,\n preconditions:x?.preconditions??null,\n failure_modes:x?.failure_modes??null,\n reusability:x?.reusability??null,\n revision:x?.revision??null,\n notes:x?.notes??null\n}}\nasync function decide(name:string,schema:any,input:any){\n const response=await openai.responses.create({\n model:\"gpt-5.6-luna\",\n instructions:[\n \"You are a generic goal-to-capability controller under a held-out behavioral test.\",\n \"Use only the runtime data provided. Do not assume skill names or hidden answers.\",\n \"Prefer current authoritative knowledge over stale/internal recall.\",\n \"Choose tools/skills from manifests, respect typed inputs, and replan after observed failure.\",\n \"Never claim success before the external verifier returns pass.\"\n ].join(\"\\n\"),\n input:JSON.stringify(input),\n text:{format:{type:\"json_schema\",name,schema}}\n });\n const out=JSON.parse(response.output_text);\n transcript.push({kind:\"MODEL_DECISION\",name,input_summary:Object.keys(input),out});\n return out;\n}\nfunction hrefAllowed(idx:any,href:string){return (idx?.sources||[]).some((x:any)=>x.href===href)}\nfunction skillIds(search:any){return (search?.results||[]).map((x:any)=>String(x.id))}\nasync function fetchManifests(base:string,idKey:string,id:string,ids:string[]){\n const out:any[]=[];\n for(const sid of ids.slice(0,12)){\n const r=await req(base,\"/skill?\"+idKey+\"=\"+encodeURIComponent(id)+\"&skill=\"+encodeURIComponent(sid));\n if(r.status===200)out.push({id:sid,manifest:compactManifest(r.json)});\n }\n return out;\n}\nconst sourceDecisionSchema={\n type:\"object\",properties:{\n knowledge_href:{type:\"string\"},\n skill_search_query:{type:\"string\"},\n reason:{type:\"string\"}\n },required:[\"knowledge_href\",\"skill_search_query\",\"reason\"],additionalProperties:false\n};\nconst planSchema={\n type:\"object\",properties:{\n ordered_skill_ids:{type:\"array\",minItems:1,items:{type:\"string\"}},\n initial_input_json:{type:\"string\"},\n retention:{type:\"array\",items:{type:\"object\",properties:{\n skill_id:{type:\"string\"},disposition:{type:\"string\",enum:[\"RETAIN\",\"CONDITIONAL\"]}\n },required:[\"skill_id\",\"disposition\"],additionalProperties:false}},\n reason:{type:\"string\"}\n },required:[\"ordered_skill_ids\",\"initial_input_json\",\"retention\",\"reason\"],additionalProperties:false\n};\nconst replanSchema={\n type:\"object\",properties:{\n replacement_search_query:{type:\"string\"},reason:{type:\"string\"}\n },required:[\"replacement_search_query\",\"reason\"],additionalProperties:false\n};\nconst replacementSchema={\n type:\"object\",properties:{\n replacement_skill_id:{type:\"string\"},reason:{type:\"string\"}\n },required:[\"replacement_skill_id\",\"reason\"],additionalProperties:false\n};\nconst revisionSchema={\n type:\"object\",properties:{\n retained:{type:\"array\",items:{type:\"string\"}},\n demoted_or_revised:{type:\"array\",items:{type:\"string\"}},\n reason:{type:\"string\"}\n },required:[\"retained\",\"demoted_or_revised\",\"reason\"],additionalProperties:false\n};\n\nasync function tokenDomain(){\n const n=await post(H_TOKEN,\"/new\",{});assert(n.status===201,\"TOKEN_NEW_FAIL\");\n const id=String(n.json.challenge_id);\n const idx=await req(H_TOKEN,\"/knowledge-index?id=\"+encodeURIComponent(id));assert(idx.status===200,\"TOKEN_INDEX_FAIL\");\n const d1=await decide(\"token_source_choice_v1\",sourceDecisionSchema,{\n goal:\"Transform the runtime input into an externally verified terminal artifact. Acquire current knowledge and runtime skills instead of relying on pretraining.\",\n runtime_input:n.json.runtime_input,knowledge_index:idx.json\n });\n assert(hrefAllowed(idx.json,d1.knowledge_href),\"TOKEN_SOURCE_NOT_FROM_INDEX\");\n const knowledge=await req(H_TOKEN,d1.knowledge_href);assert(knowledge.status===200,\"TOKEN_KNOWLEDGE_FAIL\");\n const search=await req(H_TOKEN,\"/skill-search?id=\"+encodeURIComponent(id)+\"&q=\"+encodeURIComponent(d1.skill_search_query));assert(search.status===200,\"TOKEN_SEARCH_FAIL\");\n const ids=skillIds(search.json);const manifests=await fetchManifests(H_TOKEN,\"id\",id,ids);\n const plan=await decide(\"token_skill_plan_v1\",planSchema,{\n goal:\"Compose discovered skills to transform runtime input into a terminal artifact. Use the selected current knowledge. If a skill later fails, the runtime will give you a chance to replan.\",\n runtime_input:n.json.runtime_input,knowledge:knowledge.json,skill_search_results:search.json.results,manifests\n });\n assert(plan.ordered_skill_ids.every((x:string)=>ids.includes(x)),\"TOKEN_PLAN_UNDISCOVERED_SKILL\");\n const proposedInitial:any=JSON.parse(plan.initial_input_json);\n const firstManifest:any=manifests.find((x:any)=>x.id===plan.ordered_skill_ids[0])?.manifest??null;\n const materializedInitial:any={...n.json.runtime_input,...proposedInitial,kind:n.json.runtime_input.kind};\n // Deterministically satisfy named manifest args from equally named current authoritative\n // knowledge fields when the model omitted them. This is generic schema grounding, not\n // benchmark-specific logic, and leaves unresolved args fail-closed.\n for(const arg of (Array.isArray(firstManifest?.args)?firstManifest.args:[])){\n if(!(arg in materializedInitial)&&knowledge?.claim&&Object.prototype.hasOwnProperty.call(knowledge.claim,arg)){\n materializedInitial[arg]=knowledge.claim[arg];\n }\n }\n let current:any=materializedInitial, lastGood:any=null, failed:any=null, failIndex=-1;\n lastGood=current;\n const events:any[]=[];\n for(let i=0;i=200&&r.status<300){current=r.json.output;lastGood=current;continue}\n failed={skill_id:sid,status:r.status,result:r.json};failIndex=i;break;\n }\n if(failed){\n const rp=await decide(\"token_failure_replan_v1\",replanSchema,{\n goal:\"Recover from the observed runtime skill failure without restarting or using the verifier as an oracle.\",\n selected_knowledge:knowledge.json,failed,last_good_output:lastGood,prior_plan:plan,events\n });\n const s2=await req(H_TOKEN,\"/skill-search?id=\"+encodeURIComponent(id)+\"&q=\"+encodeURIComponent(rp.replacement_search_query));assert(s2.status===200,\"TOKEN_REPLAN_SEARCH_FAIL\");\n const ids2=skillIds(s2.json);const m2=await fetchManifests(H_TOKEN,\"id\",id,ids2);\n const pick=await decide(\"token_replacement_choice_v1\",replacementSchema,{\n failure:failed,last_good_output:lastGood,search_results:s2.json.results,manifests:m2\n });\n assert(ids2.includes(pick.replacement_skill_id),\"TOKEN_REPLACEMENT_UNDISCOVERED\");\n const rr=await post(H_TOKEN,\"/execute-skill\",{challenge_id:id,skill_id:pick.replacement_skill_id,input:lastGood});\n events.push({skill_id:pick.replacement_skill_id,status:rr.status,result:rr.json,replacement:true});assert(rr.status===200,\"TOKEN_REPLACEMENT_EXEC_FAIL\");\n current=rr.json.output;\n for(let i=failIndex+1;iids.includes(x)),\"ROUTE_PLAN_UNDISCOVERED\");\n const proposedInitial:any=JSON.parse(plan.initial_input_json);\n let current:any={...n.json.runtime_input,...proposedInitial,kind:n.json.runtime_input.kind};const events1:any[]=[];\n for(const sid of plan.ordered_skill_ids){\n const r=await post(H_ROUTE,\"/execute-skill\",{session_id:id,skill_id:sid,input:current});\n events1.push({skill_id:sid,status:r.status,result:r.json});assert(r.status===200,\"ROUTE_PHASE1_SKILL_FAIL\");\n current=r.json.output;\n }\n const verify1=await post(H_ROUTE,\"/verify-phase1\",{session_id:id,artifact:current});assert(verify1.status===200&&verify1.json?.pass===true,\"ROUTE_VERIFY1_FAIL\");\n const retained=plan.retention.filter((x:any)=>x.disposition===\"RETAIN\").map((x:any)=>x.skill_id);\n const conditional=plan.retention.filter((x:any)=>x.disposition===\"CONDITIONAL\").map((x:any)=>x.skill_id);\n const retention1={schema:\"VAL_NATIVE_CANDIDATE_RETENTION_PHASE1_V1\",candidate_id:candidateId,run_id:runId,session_id:id,retained_verified:retained,conditional_or_scoped:conditional,evidence:{knowledge_href:d1.knowledge_href,skill_events:events1.map(x=>({skill_id:x.skill_id,status:x.status}))}};\n const key1=\"equivalence/\"+runId+\"/route-retention-phase1.json\";await blob.set(key1,JSON.stringify(retention1));\n const ref1=\"valblob:mosadek/brain-val-superhero-candidate-v1/\"+key1;\n const adv=await post(H_ROUTE,\"/advance-phase2\",{session_id:id,retention_claim:{retained_verified:retained,conditional_or_scoped:conditional,external_record_ref:ref1}});assert(adv.status===200,\"ROUTE_ADVANCE2_FAIL\");\n\n const idx2=await req(H_ROUTE,\"/knowledge-index?id=\"+encodeURIComponent(id));assert(idx2.status===200,\"ROUTE_INDEX2_FAIL\");\n const d2=await decide(\"route_source_choice_phase2_v1\",sourceDecisionSchema,{\n goal:\"Policy/preconditions changed. Re-evaluate current authoritative knowledge before reusing retained cognition. The search query should help only if a replacement becomes necessary.\",\n runtime_input:adv.json.runtime_input,knowledge_index:idx2.json,prior_retention:retention1\n });\n assert(hrefAllowed(idx2.json,d2.knowledge_href),\"ROUTE_SOURCE2_NOT_FROM_INDEX\");\n const knowledge2=await req(H_ROUTE,d2.knowledge_href);assert(knowledge2.status===200,\"ROUTE_KNOW2_FAIL\");\n\n current=adv.json.runtime_input;const events2:any[]=[];let failed:any=null,failIndex=-1,lastGood:any=current;\n for(let i=0;ix.status===200).map((x:any)=>String(x.skill_id)));\n const canonicalRetained=[\n ...retention1.retained_verified.filter((sid:string)=>successfulPhase2.has(sid)),\n ...(replacementId&&successfulPhase2.has(replacementId)?[replacementId]:[])\n ].filter((sid:string,i:number,a:string[])=>a.indexOf(sid)===i);\n const canonicalDemoted=failed?.skill_id?[String(failed.skill_id)]:[];\n const retention2={schema:\"VAL_NATIVE_CANDIDATE_RETENTION_PHASE2_V1\",candidate_id:candidateId,run_id:runId,session_id:id,retained:canonicalRetained,demoted_or_revised:canonicalDemoted,model_revision_reasoning:revision,evidence:{knowledge_href:d2.knowledge_href,events:events2.map(x=>({skill_id:x.skill_id,status:x.status,replacement:!!x.replacement}))}};\n const key2=\"equivalence/\"+runId+\"/route-retention-phase2.json\";await blob.set(key2,JSON.stringify(retention2));\n const ref2=\"valblob:mosadek/brain-val-superhero-candidate-v1/\"+key2;\n const verify2=await post(H_ROUTE,\"/verify-phase2\",{session_id:id,artifact:current,revision_record:{retained:canonicalRetained,demoted_or_revised:canonicalDemoted,external_record_ref:ref2}});\n const audit=await req(H_ROUTE,\"/audit?id=\"+encodeURIComponent(id));\n const rb1=await blob.get(key1).then(r=>r.text());const rb2=await blob.get(key2).then(r=>r.text());\n return{domain:\"ROUTE\",session_id:id,knowledge_phase1:d1.knowledge_href,knowledge_phase2:d2.knowledge_href,plan,events1,verify1:verify1.json,retention1_ref:ref1,events2,replacement_skill_id:replacementId,revision,verify2:verify2.json,verify2_status:verify2.status,audit:audit.json,durable_readback:JSON.parse(rb1).session_id===id&&JSON.parse(rb2).session_id===id,pass:verify2.status===200&&verify2.json?.pass===true};\n}\n\nconst started=new Date().toISOString();\nlet token:any=null,route:any=null,error:any=null;\ntry{\n token=await tokenDomain();\n assert(token.pass,\"TOKEN_DOMAIN_NOT_VERIFIED\");\n route=await routeDomain();\n assert(route.pass,\"ROUTE_DOMAIN_NOT_VERIFIED\");\n}catch(e){error=e instanceof Error?e.message:String(e)}\nconst result={\n schema:\"VAL_NATIVE_GENERIC_SUPERHERO_EQUIVALENCE_V2_CANDIDATE_RUN_V1\",\n classification:\"CANDIDATE_NONAUTHORITY_BEHAVIOR_EVIDENCE\",\n candidate_id:candidateId,run_id:runId,started_at:started,completed_at:new Date().toISOString(),\n controller_constraints:{\n hidden_harness_source_inspected_by_runtime_controller:false,\n benchmark_skill_ids_hardcoded_in_source:false,\n hidden_answer_or_mapping_hardcoded:false,\n decisions_from_public_runtime_metadata_and_model:true,\n model:\"gpt-5.6-luna\"\n },\n token_domain:token,route_domain:route,error,\n candidate_pass:error===null&&token?.pass===true&&route?.pass===true,\n transcript,\n p_real_increment:0,authority:false,independent_review_claimed:false,human_authority_approval:false,direct_sprites_calls:0,additional_out_of_pocket_spend_usd:0\n};\nconst resultText=JSON.stringify(result);\nconst dg=await crypto.subtle.digest(\"SHA-256\",new TextEncoder().encode(resultText));\nconst resultSha=[...new Uint8Array(dg)].map(b=>b.toString(16).padStart(2,\"0\")).join(\"\");\nawait blob.set(\"equivalence/\"+runId+\"/candidate-result.json\",resultText);\nawait blob.setJSON(\"equivalence/LATEST.json\",{run_id:runId,result_key:\"equivalence/\"+runId+\"/candidate-result.json\",result_sha256:resultSha,candidate_pass:result.candidate_pass});\nconsole.log(JSON.stringify({schema:result.schema,candidate_id:candidateId,run_id:runId,candidate_pass:result.candidate_pass,error,result_sha256:resultSha,token_pass:token?.pass??false,route_pass:route?.pass??false,token_challenge_id:token?.challenge_id??null,route_session_id:route?.session_id??null,p_real_increment:0},null,2));","import { ValTownOpenAI } from \"https://esm.town/v/std/openai/main.ts\";\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nimport { blob as globalBlob } from \"https://esm.town/v/std/blob/global.ts\";\n\nconst H_TOKEN=\"https://mosadek--f0167112abb811f19b3f1607ee4eb77e.web.val.run\";\nconst H_ROUTE=\"https://mosadek--b1a23924abb911f1b6d51607ee4eb77e.web.val.run\";\nconst openai=new ValTownOpenAI({});\nconst candidateId=\"VAL_NATIVE_GENERIC_JIT_CONTROLLER_V1R2\";\nconst runId=crypto.randomUUID();\nconst transcript:any[]=[];\n\nasync function req(base:string,path:string,init?:RequestInit){\n const r=await fetch(base+path,init);\n const text=await r.text();\n let json:any=null; try{json=JSON.parse(text)}catch{}\n transcript.push({kind:\"HTTP\",base,path,status:r.status,json});\n return{status:r.status,json,text};\n}\nasync function post(base:string,path:string,body:any){\n return req(base,path,{method:\"POST\",headers:{\"content-type\":\"application/json\"},body:JSON.stringify(body)});\n}\nfunction assert(x:any,msg:string){if(!x)throw new Error(msg)}\nfunction compactManifest(x:any){return{\n id:x?.id??x?.skill_id??null,\n name:x?.name??null,\n description:x?.description??null,\n input_kind:x?.input_kind??null,\n output_kind:x?.output_kind??null,\n input_schema:x?.input_schema??x?.inputSchema??x?.input??null,\n output_schema:x?.output_schema??x?.outputSchema??x?.output??null,\n args:x?.args??null,\n preconditions:x?.preconditions??null,\n failure_modes:x?.failure_modes??null,\n reusability:x?.reusability??null,\n revision:x?.revision??null,\n notes:x?.notes??null\n}}\nasync function decide(name:string,schema:any,input:any){\n const response=await openai.responses.create({\n model:\"gpt-5.6-luna\",\n instructions:[\n \"You are a generic goal-to-capability controller under a held-out behavioral test.\",\n \"Use only the runtime data provided. Do not assume skill names or hidden answers.\",\n \"Prefer current authoritative knowledge over stale/internal recall.\",\n \"Choose tools/skills from manifests, respect typed inputs, and replan after observed failure.\",\n \"Never claim success before the external verifier returns pass.\"\n ].join(\"\\n\"),\n input:JSON.stringify(input),\n text:{format:{type:\"json_schema\",name,schema}}\n });\n const out=JSON.parse(response.output_text);\n transcript.push({kind:\"MODEL_DECISION\",name,input_summary:Object.keys(input),out});\n return out;\n}\nfunction hrefAllowed(idx:any,href:string){return (idx?.sources||[]).some((x:any)=>x.href===href)}\nfunction skillIds(search:any){return (search?.results||[]).map((x:any)=>String(x.id))}\nasync function fetchManifests(base:string,idKey:string,id:string,ids:string[]){\n const out:any[]=[];\n for(const sid of ids.slice(0,12)){\n const r=await req(base,\"/skill?\"+idKey+\"=\"+encodeURIComponent(id)+\"&skill=\"+encodeURIComponent(sid));\n if(r.status===200)out.push({id:sid,manifest:compactManifest(r.json)});\n }\n return out;\n}\nconst sourceDecisionSchema={\n type:\"object\",properties:{\n knowledge_href:{type:\"string\"},\n skill_search_query:{type:\"string\"},\n reason:{type:\"string\"}\n },required:[\"knowledge_href\",\"skill_search_query\",\"reason\"],additionalProperties:false\n};\nconst planSchema={\n type:\"object\",properties:{\n ordered_skill_ids:{type:\"array\",minItems:1,items:{type:\"string\"}},\n initial_input_json:{type:\"string\"},\n retention:{type:\"array\",items:{type:\"object\",properties:{\n skill_id:{type:\"string\"},disposition:{type:\"string\",enum:[\"RETAIN\",\"CONDITIONAL\"]}\n },required:[\"skill_id\",\"disposition\"],additionalProperties:false}},\n reason:{type:\"string\"}\n },required:[\"ordered_skill_ids\",\"initial_input_json\",\"retention\",\"reason\"],additionalProperties:false\n};\nconst replanSchema={\n type:\"object\",properties:{\n replacement_search_query:{type:\"string\"},reason:{type:\"string\"}\n },required:[\"replacement_search_query\",\"reason\"],additionalProperties:false\n};\nconst replacementSchema={\n type:\"object\",properties:{\n replacement_skill_id:{type:\"string\"},reason:{type:\"string\"}\n },required:[\"replacement_skill_id\",\"reason\"],additionalProperties:false\n};\nconst revisionSchema={\n type:\"object\",properties:{\n retained:{type:\"array\",items:{type:\"string\"}},\n demoted_or_revised:{type:\"array\",items:{type:\"string\"}},\n reason:{type:\"string\"}\n },required:[\"retained\",\"demoted_or_revised\",\"reason\"],additionalProperties:false\n};\n\nasync function tokenDomain(){\n const n=await post(H_TOKEN,\"/new\",{});assert(n.status===201,\"TOKEN_NEW_FAIL\");\n const id=String(n.json.challenge_id);\n const idx=await req(H_TOKEN,\"/knowledge-index?id=\"+encodeURIComponent(id));assert(idx.status===200,\"TOKEN_INDEX_FAIL\");\n const d1=await decide(\"token_source_choice_v1\",sourceDecisionSchema,{\n goal:\"Transform the runtime input into an externally verified terminal artifact. Acquire current knowledge and runtime skills instead of relying on pretraining.\",\n runtime_input:n.json.runtime_input,knowledge_index:idx.json\n });\n assert(hrefAllowed(idx.json,d1.knowledge_href),\"TOKEN_SOURCE_NOT_FROM_INDEX\");\n const knowledge=await req(H_TOKEN,d1.knowledge_href);assert(knowledge.status===200,\"TOKEN_KNOWLEDGE_FAIL\");\n const search=await req(H_TOKEN,\"/skill-search?id=\"+encodeURIComponent(id)+\"&q=\"+encodeURIComponent(d1.skill_search_query));assert(search.status===200,\"TOKEN_SEARCH_FAIL\");\n const ids=skillIds(search.json);const manifests=await fetchManifests(H_TOKEN,\"id\",id,ids);\n const plan=await decide(\"token_skill_plan_v1\",planSchema,{\n goal:\"Compose discovered skills to transform runtime input into a terminal artifact. Use the selected current knowledge. If a skill later fails, the runtime will give you a chance to replan.\",\n runtime_input:n.json.runtime_input,knowledge:knowledge.json,skill_search_results:search.json.results,manifests\n });\n assert(plan.ordered_skill_ids.every((x:string)=>ids.includes(x)),\"TOKEN_PLAN_UNDISCOVERED_SKILL\");\n const proposedInitial:any=JSON.parse(plan.initial_input_json);\n const firstManifest:any=manifests.find((x:any)=>x.id===plan.ordered_skill_ids[0])?.manifest??null;\n const materializedInitial:any={...n.json.runtime_input,...proposedInitial,kind:n.json.runtime_input.kind};\n // Deterministically satisfy named manifest args from equally named current authoritative\n // knowledge fields when the model omitted them. This is generic schema grounding, not\n // benchmark-specific logic, and leaves unresolved args fail-closed.\n for(const arg of (Array.isArray(firstManifest?.args)?firstManifest.args:[])){\n if(!(arg in materializedInitial)&&knowledge?.claim&&Object.prototype.hasOwnProperty.call(knowledge.claim,arg)){\n materializedInitial[arg]=knowledge.claim[arg];\n }\n }\n let current:any=materializedInitial, lastGood:any=null, failed:any=null, failIndex=-1;\n lastGood=current;\n const events:any[]=[];\n for(let i=0;i=200&&r.status<300){current=r.json.output;lastGood=current;continue}\n failed={skill_id:sid,status:r.status,result:r.json};failIndex=i;break;\n }\n if(failed){\n const rp=await decide(\"token_failure_replan_v1\",replanSchema,{\n goal:\"Recover from the observed runtime skill failure without restarting or using the verifier as an oracle.\",\n selected_knowledge:knowledge.json,failed,last_good_output:lastGood,prior_plan:plan,events\n });\n const s2=await req(H_TOKEN,\"/skill-search?id=\"+encodeURIComponent(id)+\"&q=\"+encodeURIComponent(rp.replacement_search_query));assert(s2.status===200,\"TOKEN_REPLAN_SEARCH_FAIL\");\n const ids2=skillIds(s2.json);const m2=await fetchManifests(H_TOKEN,\"id\",id,ids2);\n const pick=await decide(\"token_replacement_choice_v1\",replacementSchema,{\n failure:failed,last_good_output:lastGood,search_results:s2.json.results,manifests:m2\n });\n assert(ids2.includes(pick.replacement_skill_id),\"TOKEN_REPLACEMENT_UNDISCOVERED\");\n const rr=await post(H_TOKEN,\"/execute-skill\",{challenge_id:id,skill_id:pick.replacement_skill_id,input:lastGood});\n events.push({skill_id:pick.replacement_skill_id,status:rr.status,result:rr.json,replacement:true});assert(rr.status===200,\"TOKEN_REPLACEMENT_EXEC_FAIL\");\n current=rr.json.output;\n for(let i=failIndex+1;iids.includes(x)),\"ROUTE_PLAN_UNDISCOVERED\");\n const proposedInitial:any=JSON.parse(plan.initial_input_json);\n let current:any={...n.json.runtime_input,...proposedInitial,kind:n.json.runtime_input.kind};const events1:any[]=[];\n for(const sid of plan.ordered_skill_ids){\n const r=await post(H_ROUTE,\"/execute-skill\",{session_id:id,skill_id:sid,input:current});\n events1.push({skill_id:sid,status:r.status,result:r.json});assert(r.status===200,\"ROUTE_PHASE1_SKILL_FAIL\");\n current=r.json.output;\n }\n const verify1=await post(H_ROUTE,\"/verify-phase1\",{session_id:id,artifact:current});assert(verify1.status===200&&verify1.json?.pass===true,\"ROUTE_VERIFY1_FAIL\");\n const retained=plan.retention.filter((x:any)=>x.disposition===\"RETAIN\").map((x:any)=>x.skill_id);\n const conditional=plan.retention.filter((x:any)=>x.disposition===\"CONDITIONAL\").map((x:any)=>x.skill_id);\n const retention1={schema:\"VAL_NATIVE_CANDIDATE_RETENTION_PHASE1_V1\",candidate_id:candidateId,run_id:runId,session_id:id,retained_verified:retained,conditional_or_scoped:conditional,evidence:{knowledge_href:d1.knowledge_href,skill_events:events1.map(x=>({skill_id:x.skill_id,status:x.status}))}};\n const key1=\"equivalence/\"+runId+\"/route-retention-phase1.json\";await blob.set(key1,JSON.stringify(retention1));\n const ref1=\"valblob:mosadek/brain-val-superhero-candidate-v1/\"+key1;\n const adv=await post(H_ROUTE,\"/advance-phase2\",{session_id:id,retention_claim:{retained_verified:retained,conditional_or_scoped:conditional,external_record_ref:ref1}});assert(adv.status===200,\"ROUTE_ADVANCE2_FAIL\");\n\n const idx2=await req(H_ROUTE,\"/knowledge-index?id=\"+encodeURIComponent(id));assert(idx2.status===200,\"ROUTE_INDEX2_FAIL\");\n const d2=await decide(\"route_source_choice_phase2_v1\",sourceDecisionSchema,{\n goal:\"Policy/preconditions changed. Re-evaluate current authoritative knowledge before reusing retained cognition. The search query should help only if a replacement becomes necessary.\",\n runtime_input:adv.json.runtime_input,knowledge_index:idx2.json,prior_retention:retention1\n });\n assert(hrefAllowed(idx2.json,d2.knowledge_href),\"ROUTE_SOURCE2_NOT_FROM_INDEX\");\n const knowledge2=await req(H_ROUTE,d2.knowledge_href);assert(knowledge2.status===200,\"ROUTE_KNOW2_FAIL\");\n\n current=adv.json.runtime_input;const events2:any[]=[];let failed:any=null,failIndex=-1,lastGood:any=current;\n for(let i=0;ix.status===200).map((x:any)=>String(x.skill_id)));\n const canonicalRetained=[\n ...retention1.retained_verified.filter((sid:string)=>successfulPhase2.has(sid)),\n ...(replacementId&&successfulPhase2.has(replacementId)?[replacementId]:[])\n ].filter((sid:string,i:number,a:string[])=>a.indexOf(sid)===i);\n const canonicalDemoted=failed?.skill_id?[String(failed.skill_id)]:[];\n const retention2={schema:\"VAL_NATIVE_CANDIDATE_RETENTION_PHASE2_V1\",candidate_id:candidateId,run_id:runId,session_id:id,retained:canonicalRetained,demoted_or_revised:canonicalDemoted,model_revision_reasoning:revision,evidence:{knowledge_href:d2.knowledge_href,events:events2.map(x=>({skill_id:x.skill_id,status:x.status,replacement:!!x.replacement}))}};\n const key2=\"equivalence/\"+runId+\"/route-retention-phase2.json\";await blob.set(key2,JSON.stringify(retention2));\n const ref2=\"valblob:mosadek/brain-val-superhero-candidate-v1/\"+key2;\n const verify2=await post(H_ROUTE,\"/verify-phase2\",{session_id:id,artifact:current,revision_record:{retained:canonicalRetained,demoted_or_revised:canonicalDemoted,external_record_ref:ref2}});\n const audit=await req(H_ROUTE,\"/audit?id=\"+encodeURIComponent(id));\n const rb1=await blob.get(key1).then(r=>r.text());const rb2=await blob.get(key2).then(r=>r.text());\n return{domain:\"ROUTE\",session_id:id,knowledge_phase1:d1.knowledge_href,knowledge_phase2:d2.knowledge_href,plan,events1,verify1:verify1.json,retention1_ref:ref1,events2,replacement_skill_id:replacementId,revision,verify2:verify2.json,verify2_status:verify2.status,audit:audit.json,durable_readback:JSON.parse(rb1).session_id===id&&JSON.parse(rb2).session_id===id,pass:verify2.status===200&&verify2.json?.pass===true};\n}\n\nconst started=new Date().toISOString();\nlet token:any=null,route:any=null,error:any=null;\ntry{\n token=await tokenDomain();\n assert(token.pass,\"TOKEN_DOMAIN_NOT_VERIFIED\");\n route=await routeDomain();\n assert(route.pass,\"ROUTE_DOMAIN_NOT_VERIFIED\");\n}catch(e){error=e instanceof Error?e.message:String(e)}\nconst result={\n schema:\"VAL_NATIVE_GENERIC_SUPERHERO_EQUIVALENCE_V2_CANDIDATE_RUN_V1\",\n classification:\"CANDIDATE_NONAUTHORITY_BEHAVIOR_EVIDENCE\",\n candidate_id:candidateId,run_id:runId,started_at:started,completed_at:new Date().toISOString(),\n controller_constraints:{\n hidden_harness_source_inspected_by_runtime_controller:false,\n benchmark_skill_ids_hardcoded_in_source:false,\n hidden_answer_or_mapping_hardcoded:false,\n decisions_from_public_runtime_metadata_and_model:true,\n model:\"gpt-5.6-luna\"\n },\n token_domain:token,route_domain:route,error,\n candidate_pass:error===null&&token?.pass===true&&route?.pass===true,\n transcript,\n p_real_increment:0,authority:false,independent_review_claimed:false,human_authority_approval:false,direct_sprites_calls:0,additional_out_of_pocket_spend_usd:0\n};\nconst resultText=JSON.stringify(result);\nconst dg=await crypto.subtle.digest(\"SHA-256\",new TextEncoder().encode(resultText));\nconst resultSha=[...new Uint8Array(dg)].map(b=>b.toString(16).padStart(2,\"0\")).join(\"\");\nawait blob.set(\"equivalence/\"+runId+\"/candidate-result.json\",resultText);\nawait blob.setJSON(\"equivalence/LATEST.json\",{run_id:runId,result_key:\"equivalence/\"+runId+\"/candidate-result.json\",result_sha256:resultSha,candidate_pass:result.candidate_pass});\nconsole.log(JSON.stringify({schema:result.schema,candidate_id:candidateId,run_id:runId,candidate_pass:result.candidate_pass,error,result_sha256:resultSha,token_pass:token?.pass??false,route_pass:route?.pass??false,token_challenge_id:token?.challenge_id??null,route_session_id:route?.session_id??null,p_real_increment:0},null,2));","createdAt",["D",1788896877184],"branchVersion","branchId","3cf221c2-bc87-477b-9b99-774209340c45","id","bb689585-eecb-442c-97c7-7fafc49fe348","brain-val-superhero-candidate-v1",["D",1788896582199],"deletedAt","legacyEnvVars","privacy","public","httpPrivacy","description","Generic held-out superhero equivalence candidate","imageUrl","6be0fc5c-1fef-4bd6-a1e1-67a3f8c0764e","pinnedFileId","71dce240-abc2-11f1-9482-1607ee4eb77e","pinPreview",true,"pinMigrated","screenshotVersion","likeCount",0,"orgId","01a0715b-703f-741f-ab35-c85020e4a5e0","tursoName","bb689585eecb442c97c77fafc49fe348","tursoHostname","bb689585eecb442c97c77fafc49fe348-val-town.aws-us-east-2.turso.io","tursoMeasuredAt",["D",1789563628475],"pinned","userPerm","r","valPerm","blobPerm","rw","emailPerm","sqlitePerm","projectPerm","telemetryPerm","user",{"_100":99,"_108":104},"org",{"_100":99,"_101":102,"_103":104,"_18":55,"_105":106,"_107":25},"branch",{"_13":89,"_32":90,"_43":42,"_91":44,"_92":-5,"_93":-5,"_94":-5,"_95":-5,"_96":-5,"_97":45,"_98":99},"isOwned",61,"main","projectId","forkedBranchId","forkedBranchVersion","forkedProjectPrivacy","forkedProjectId","forkedProjectOrgId","projectName","userHandle","mosadek","handle","displayName","Mo Sadek","avatar","https://lh3.googleusercontent.com/a/ACg8ocL1eT_LHhKSg5px3nXIO6eez79aZXef466qXKh2gak-0iwrvw=s96-c","tier","free","canRestrictVals","profileImageUrl","branchName","","adoptOrgId",{"_43":44,"_32":45,"_38":155,"_47":-5,"_48":25,"_49":50,"_51":50,"_52":53,"_54":-5,"_18":55,"_56":57,"_58":59,"_60":-5,"_61":-5,"_62":63,"_64":65,"_66":67,"_68":69,"_70":156,"_72":-5,"_73":74,"_75":74,"_76":77,"_78":77,"_79":77,"_80":74,"_81":74,"_82":157,"_84":158,"_86":159,"_88":25},"tree",[150],"files",[150],"counts",{"_139":63,"_140":141,"_142":63,"_143":144,"_145":146},"contents",[131],"upstreamProject","upstreamBranch","emptyEnvVars","flags",[],"hasHttpFiles","grants","orgs","customDomains",{},{"_43":132,"_32":28,"_30":31,"_133":-5,"_41":42,"_40":14,"_134":25,"_38":135,"_13":136,"_33":34,"_35":37,"_24":25,"_137":-5,"_27":138},"155c6635-b07d-4e92-b655-c600426dfc0c","parentId","latest",["D",1788896877184],8,"label",[28],"pulls","branches",2,"forks","envVars",{"_147":63,"_149":63},"connections",{"_147":63,"_148":63},"count","unfilledCount","emptyCount",{"_43":132,"_32":28,"_30":31,"_133":-5,"_41":42,"_40":14,"_134":25,"_38":151,"_13":136,"_33":34,"_35":37,"_24":25,"_137":-5,"_152":153,"_27":154},["D",1788896877184],"children",[],[28],["D",1788896582199],["D",1789563628475],{"_100":99,"_108":104},{"_100":99,"_101":102,"_103":104,"_18":55,"_105":106,"_107":25},{"_13":89,"_32":90,"_43":42,"_91":44,"_92":-5,"_93":-5,"_94":-5,"_95":-5,"_96":-5,"_97":45,"_98":99},"ctx",{"_82":-5,"_128":193,"_194":195,"_196":197,"_198":25,"_199":25,"_200":25,"_201":59,"_202":-5,"_203":-7,"_204":-7,"_205":63,"_206":25,"_207":25,"_208":-5,"_209":25,"_210":25,"_211":25,"_212":-5},"ENV",{"_164":165,"_166":167,"_168":90,"_169":170,"_171":172,"_173":174,"_175":176,"_177":178,"_179":180,"_181":182,"_183":184,"_185":186,"_187":188,"_189":190,"_191":192},"REACT_APP_SENTRY_DSN","https://586020d86f05489db63a554690acf1e9@o1398359.ingest.sentry.io/6724499","RENDER_GIT_COMMIT","fdb67791413f4386b4dbf18f354029a617f6160f","RENDER_GIT_BRANCH","POSTHOG_API_TOKEN","phc_nmoyuAkStEZuUp69rYlg1DZgrIt19Gx0Y4Hwf8tGgJP","POSTHOG_API_HOST","https://wart.val.town","WEB_API_TEMPLATE","https://{val}.web.val.run","APP_URL","https://www.val.town","MODULE_URL","https://esm.town","VANITY_WEB_API_TEMPLATE","https://{label}.val.run/","DENOLS_URL","https://lsp.val.town","NODE_ENV","production","CLOUDFLARE_TURNSTILE_SITE_KEY","0x4AAAAAACtaIZJH7s9rBMZ6","GOOGLE_CLIENT_ID","875989696899-jtsqkhlbdv5jrlsk8631p7ipjg3c6o62.apps.googleusercontent.com","STRIPE_PUBLISHABLE_KEY","pk_live_51N2zvWEpZTTLipIkVEwXrUG9AT0yeOOEYxszarZT1jpsZV58Ly6wlg3SltZpHN26CUalOBmUSY7adQRQiSkCZz2500qSXNkf8Y","API_URL","https://api.val.town",[],"preferencesKeybindings","default","preferencesTheme","light","preferencesFixers","preferencesCopilot","preferencesSelectionAi","preferencesCloseBrackets","preferencesChatPanelState","preferencesDprint","preferencesDenofmt","notificationCount","flagDenols","isBusinessOrgMember","businessOrgHandle","isBusinessOrgAdmin","isHigherTownieTier","navCollapsed","currentOrgId"]

Diff · v9