alchemylab 1.4.12 → 1.4.14

package/dist/cli.js

@@ -946,18 +946,28 @@ module.exports=require("ws")},5124(module){"use strict"
 module.exports=require("winston")},5317(module){"use strict"
 module.exports=require("child_process")},5419(e,exports,t){"use strict"
 function r(e){return(0,o.join)((0,s.homedir)(),n.ALCHEMY_DIR_NAME,e?"opencode-dev":"opencode")}Object.defineProperty(exports,"__esModule",{value:!0}),exports.getGlobalConfigDir=r,exports.getOpencodeSpawnEnv=function(e,t){return{OPENCODE_CLIENT:"alchemy",OPENCODE_DATA_DIR:(0,o.join)(e,n.ALCHEMY_DIR_NAME,"opencode"),OPENCODE_CONFIG_DIR:r(t)}}
-const s=t(857),o=t(6928),n=t(4139)},5607(e,exports,t){"use strict"
-Object.defineProperty(exports,"__esModule",{value:!0}),exports.executeAction=async function(e,t,r){const s=f[e]
+const s=t(857),o=t(6928),n=t(4139)},5454(e,exports){"use strict"
+function t(e,t){let r=0
+for(let t=0;t<e.length;t++)r=(r<<5)-r+e.charCodeAt(t)|0
+return`${t+1}${String.fromCharCode(97+Math.abs(r)%26)}`}function r(e){const r=new Map
+for(let s=0;s<e.length;s++){const o=t(e[s].id,s)
+r.set(o,e[s])}return r}function s(e){if(0===e.length)return"No acceptance criteria defined."
+const t=r(e),s=[]
+for(const[e,r]of t){const t=(r.status||"unverified").toUpperCase()
+s.push(`[${e}] [${t}] ${r.description}`)}return s.join("\n")}function o(e){const t=r(e),o=[]
+for(const[e,r]of t)o.push(`    {"key": "${e}", "evidence": "<your evidence for: ${r.description}>"}`)
+return`Use the merge operation with evidence for each criterion:\n\n${s(e)}\n\nExample call:\n{\n  "operation": "merge",\n  "requestId": "<requestId>",\n  "merge": {\n    "verifications": [\n${o.join(",\n")}\n    ]\n  }\n}`}Object.defineProperty(exports,"__esModule",{value:!0}),exports.formatEvidenceInstructions=exports.formatCriteriaWithHashes=exports.buildCriteriaHashMap=exports.criterionHash=void 0,exports.criterionShortKey=t,exports.buildCriteriaKeyMap=r,exports.formatCriteriaWithKeys=s,exports.formatVerifyInstructions=o,exports.criterionHash=t,exports.buildCriteriaHashMap=r,exports.formatCriteriaWithHashes=s,exports.formatEvidenceInstructions=o},5607(e,exports,t){"use strict"
+Object.defineProperty(exports,"__esModule",{value:!0}),exports.executeAction=async function(e,t,r){const s=y[e]
 if(!s)return{success:!1,message:"Unknown action type: "+e,error:"No handler registered for action type: "+e}
-try{return await s(t,r)}catch(t){const s=(0,p.getErrorMessage)(t)
+try{return await s(t,r)}catch(t){const s=(0,m.getErrorMessage)(t)
 return r.logger.error("Action handler failed",{actionType:e,error:s}),{success:!1,message:"Action failed: "+s,error:s}}}
-const r=t(6636),s=t(7453),o=t(5700),n=t(848),i=t(9735),a=t(7767),c=t(9266),l=t(6555),d=t(3614),u=t(4654),h=t(7078),g=t(1405),p=t(8824),m=t(2272),f={revert_request:r.handleRevertRequestAction,"revert-request":r.handleRevertRequestAction,complete_request:s.handleCompleteRequestAction,"complete-request":s.handleCompleteRequestAction,"finalize-request":n.handleFinalizeRequestAction,"approve-request":i.handleApproveRequestAction,"approve-and-work-request":a.handleApproveAndWorkRequestAction,"restart-mcp":o.handleRestartMcpAction,"opencode-proxy":c.handleOpenCodeProxyAction,"run-tests":l.handleRunTestsAction,async"inject-prompt"(e,r){const{sessionId:s,message:o,kind:n,partPayload:i}=e
+const r=t(6636),s=t(7453),o=t(5700),n=t(848),i=t(9735),a=t(7767),c=t(9266),l=t(6555),d=t(3614),u=t(4654),h=t(7078),g=t(7775),p=t(1405),m=t(8824),f=t(2272),y={revert_request:r.handleRevertRequestAction,"revert-request":r.handleRevertRequestAction,complete_request:s.handleCompleteRequestAction,"complete-request":s.handleCompleteRequestAction,"finalize-request":n.handleFinalizeRequestAction,"approve-request":i.handleApproveRequestAction,"approve-and-work-request":a.handleApproveAndWorkRequestAction,"restart-mcp":o.handleRestartMcpAction,"opencode-proxy":c.handleOpenCodeProxyAction,"run-tests":l.handleRunTestsAction,async"inject-prompt"(e,r){const{sessionId:s,message:o,kind:n,partPayload:i}=e
 if(!s||!o)return{success:!1,message:"sessionId and message are required",error:"missing_params"}
 const{getOpenCodeUrlFromConfig:a}=await Promise.resolve().then(t.t.bind(t,833,23)),c=a(r.projectPath)
 if(!c)return{success:!1,message:"Cannot read OpenCode config",error:"no_opencode"}
-try{const e=await fetch(`${c}/session/${s}/prompt_async`,{method:"POST",headers:{"Content-Type":"application/json"},body:JSON.stringify({parts:[{type:"text",text:o,metadata:(0,m.partMeta)(n??"inserted-prompt",i)}],...r.cachedDefaultModel?(()=>{const[e,...t]=r.cachedDefaultModel.split("/")
+try{const e=await fetch(`${c}/session/${s}/prompt_async`,{method:"POST",headers:{"Content-Type":"application/json"},body:JSON.stringify({parts:[{type:"text",text:o,metadata:(0,f.partMeta)(n??"inserted-prompt",i)}],...r.cachedDefaultModel?(()=>{const[e,...t]=r.cachedDefaultModel.split("/")
 return{model:{providerID:e,modelID:t.join("/")}}})():{}}),signal:AbortSignal.timeout(1e4)})
-return e.ok?{success:!0,message:"Prompt injected into session "+s}:{success:!1,message:"prompt_async failed: "+e.status,error:"prompt_failed"}}catch(e){return{success:!1,message:(0,p.getErrorMessage)(e),error:"prompt_error"}}},"fork-filesystem":d.handleForkFilesystemAction,"prompt-and-wait":u.handlePromptAndWaitAction,"run-ab-prompt":u.handlePromptAndWaitAction,"spawn-auto-planner":h.handleSpawnAutoPlannerAction,"start-building":g.handleStartBuildingAction}},5671(e,exports,t){"use strict"
+return e.ok?{success:!0,message:"Prompt injected into session "+s}:{success:!1,message:"prompt_async failed: "+e.status,error:"prompt_failed"}}catch(e){return{success:!1,message:(0,m.getErrorMessage)(e),error:"prompt_error"}}},"fork-filesystem":d.handleForkFilesystemAction,"prompt-and-wait":u.handlePromptAndWaitAction,"run-ab-prompt":u.handlePromptAndWaitAction,"spawn-auto-planner":h.handleSpawnAutoPlannerAction,"spawn-reviewer":g.handleSpawnReviewerAction,"start-building":p.handleStartBuildingAction}},5671(e,exports,t){"use strict"
 function r(e){const[t,...r]=e.split("/")
 return{provider:t||"anthropic",modelId:r.join("/")||e}}function s(e){const t={anthropic:!1,openai:!1}
 try{const s=(0,a.join)(e,"opencode.json")
@@ -1036,35 +1046,33 @@ return"idle"===t||"completed"===t}))
 return{sessions:s,statusMap:o,running:n,idle:i}}catch(e){return this.logger.debug("Failed to fetch OpenCode session state",{error:e+""}),null}}}},6540(e,exports){"use strict"
 var t
 Object.defineProperty(exports,"__esModule",{value:!0}),exports.LogLevel=void 0,function(e){e.ERROR="error",e.WARN="warn",e.INFO="info",e.DEBUG="debug"}(t||(exports.LogLevel=t={}))},6555(e,exports,t){"use strict"
-function r(e,process,t,r,s=!1,o,n){const i=process.config||{},a=(i.setup,i.commands||[]),c=o&&o.length>0?o.map((e=>a.find((t=>t.name===e)))).filter((e=>e)):a,l=new Set(n||[]),d=l.size>0?c.filter((e=>!l.has(e.name))):c,u=o&&o.length>0&&d.length<a.length||l.size>0,h=l.size>0?a.filter((e=>l.has(e.name))).map((e=>e.name)):[]
-let g=`You are a test runner agent. Your job is to verify that request ${e.localId||e.id} ("${e.title}") passes all required tests.\n\n## Test Process: ${process.name}\n${process.description?`\n${process.description}\n`:""}\n${process.scope?`**Scope:** Files under \`${process.scope}\`\n`:""}\n## Project ID: ${r}\n${h.length>0?`\n_The coder opted out of these command(s) for this completion: **${h.join(", ")}**. Do NOT run them. They've been excluded from the list below._\n`:""}\n## Test Commands${u?" (partial run \u2014 only the commands listed below)":""}\nExecute these in order:\n`
-for(let e=0;e<d.length;e++){const t=d[e]
-g+=`\n### ${e+1}. ${t.name}${!1!==t.required?" (REQUIRED)":" (optional)"}\n- **Command:** \`${t.command}\`\n- **Timeout:** ${t.timeout||300}s\n`}if(g+=s?`\n## Instructions\n\nYou are in interactive mode. Present the test plan to the user and **wait for their direction** before running anything.\n\n1. Greet the user and list the test commands available for this process.\n2. Wait for the user to tell you to proceed (they may want to run specific commands, skip some, or run all).\n3. Use \`execute_test_command\` with processId="${process.id}" and the command name to run each command.\n4. After each command, explain the result and wait for the user before continuing.\n5. When the user is satisfied, call \`report_test_results\` with the overall status.\n\nYou are read-only \u2014 do NOT modify any source code. Your job is to verify, not fix.`:`\n## Instructions\n\nSetup commands have already been executed. Use the \`execute_test_command\` tool to run each test command:\n\n1. For each command above, call \`execute_test_command\` with processId="${process.id}" and the command name.\n2. Check if it passed or failed.\n3. If it fails: try to determine why \u2014 is it a real bug, a flaky test, or an environment issue? Retry once if it looks flaky.\n4. When all commands are done, call \`report_test_results\` with the overall status and per-command results.\n   - If all REQUIRED commands passed \u2192 status "passed"\n   - If any REQUIRED command failed \u2192 status "failed" with diagnosis of each failure\n\nYou are read-only \u2014 do NOT modify any source code. Your job is to verify, not fix.`,i.verifyAcceptanceCriteria&&e.acceptanceCriteria?.length>0){g+="\n\n## Acceptance Criteria Verification\n\nAfter running all commands, verify each acceptance criterion by reading code changes and running ad-hoc commands as needed. This is **advisory only** \u2014 it does not affect the overall pass/fail status.\n\nCriteria to verify:\n"
-for(const t of e.acceptanceCriteria)g+=`- ${t}\n`
-g+='\nFor each criterion, report in the `acceptanceCriteria` array of your results:\n- **criterion**: The criterion text\n- **status**: "verified" (confirmed in code), "unverified" (not found/not implemented), or "unclear" (ambiguous)\n- **evidence**: How you verified it, or why you couldn\'t\n'}return g}Object.defineProperty(exports,"__esModule",{value:!0}),exports.handleRunTestsAction=async function(e,l){const{requestId:d,testProcessIds:u,parentSessionId:h,mode:g}=e,p="interactive"===g,{logger:m,apiClient:f,projectId:y,projectPath:v,trackChildPid:w}=l
+function r(e,process,t,r,s=!1,o,n){const i=process.config||{},c=(i.setup,i.commands||[]),l=o&&o.length>0?o.map((e=>c.find((t=>t.name===e)))).filter((e=>e)):c,d=new Set(n||[]),u=d.size>0?l.filter((e=>!d.has(e.name))):l,h=o&&o.length>0&&u.length<c.length||d.size>0,g=d.size>0?c.filter((e=>d.has(e.name))).map((e=>e.name)):[]
+let p=`You are a test runner agent. Your job is to verify that request ${e.localId||e.id} ("${e.title}") passes all required tests.\n\n## Test Process: ${process.name}\n${process.description?`\n${process.description}\n`:""}\n${process.scope?`**Scope:** Files under \`${process.scope}\`\n`:""}\n## Project ID: ${r}\n${g.length>0?`\n_The coder opted out of these command(s) for this completion: **${g.join(", ")}**. Do NOT run them. They've been excluded from the list below._\n`:""}\n## Test Commands${h?" (partial run \u2014 only the commands listed below)":""}\nExecute these in order:\n`
+for(let e=0;e<u.length;e++){const t=u[e]
+p+=`\n### ${e+1}. ${t.name}${!1!==t.required?" (REQUIRED)":" (optional)"}\n- **Command:** \`${t.command}\`\n- **Timeout:** ${t.timeout||300}s\n`}return p+=s?`\n## Instructions\n\nYou are in interactive mode. Present the test plan to the user and **wait for their direction** before running anything.\n\n1. Greet the user and list the test commands available for this process.\n2. Wait for the user to tell you to proceed (they may want to run specific commands, skip some, or run all).\n3. Use \`execute_test_command\` with processId="${process.id}" and the command name to run each command.\n4. After each command, explain the result and wait for the user before continuing.\n5. When the user is satisfied, call \`report_test_results\` with the overall status.\n\nYou are read-only \u2014 do NOT modify any source code. Your job is to verify, not fix.`:`\n## Instructions\n\nSetup commands have already been executed. Use the \`execute_test_command\` tool to run each test command:\n\n1. For each command above, call \`execute_test_command\` with processId="${process.id}" and the command name.\n2. Check if it passed or failed.\n3. If it fails: try to determine why \u2014 is it a real bug, a flaky test, or an environment issue? Retry once if it looks flaky.\n4. When all commands are done, call \`report_test_results\` with the overall status and per-command results.\n   - If all REQUIRED commands passed \u2192 status "passed"\n   - If any REQUIRED command failed \u2192 status "failed" with diagnosis of each failure\n\nYou are read-only \u2014 do NOT modify any source code. Your job is to verify, not fix.`,i.verifyAcceptanceCriteria&&e.acceptanceCriteria?.length>0&&(p+="\n\n## Acceptance Criteria Verification\n\nAfter running all commands, verify each acceptance criterion by running the relevant tests/commands (and reading code as needed). Criteria you confirm by execution are recorded as the test-dimension verdict.\n\nCriteria to verify (each is [key] [STATUS] text \u2014 cite the key):\n",p+=(0,a.formatCriteriaWithKeys)(e.acceptanceCriteria)+"\n",p+='\nFor each criterion, report in the `acceptanceCriteria` array of your results:\n- **key**: the criterion\'s [key] shown above\n- **status**: "verified" (proven by running it), "unverified" (not satisfied), or "unclear" (couldn\'t determine)\n- **evidence**: the command/output that proves it, or why you couldn\'t\n'),p}Object.defineProperty(exports,"__esModule",{value:!0}),exports.handleRunTestsAction=async function(e,a){const{requestId:d,testProcessIds:u,parentSessionId:h,mode:g}=e,p="interactive"===g,{logger:m,apiClient:f,projectId:y,projectPath:v,trackChildPid:w}=a
 if(!d)return{success:!1,message:"requestId is required",error:"Missing requestId in payload"}
 m.info("Starting run-tests action",{requestId:d,testProcessIds:u})
-try{const l=await f.get(`/api/projects/${y}/requests/${d}`)
-if(!l.success||!l.data)return{success:!1,message:`Request ${d} not found`,error:"Request not found"}
-const g=l.data,I=g.branchName||"requests/"+d
+try{const a=await f.get(`/api/projects/${y}/requests/${d}`)
+if(!a.success||!a.data)return{success:!1,message:`Request ${d} not found`,error:"Request not found"}
+const g=a.data,I=g.branchName||"requests/"+d
 let E
 if(u?.length)E=(await Promise.all(u.map((e=>f.get(`/api/projects/${y}/test-processes/${e}`))))).filter((e=>e.success&&e.data)).map((e=>e.data))
 else{const e=await f.get(`/api/projects/${y}/test-processes/active`)
 E=e.success&&e.data?Array.isArray(e.data)?e.data:[e.data]:[]}if(0===E.length)return{success:!1,message:"No test processes configured for this project. Set up testing first.",error:"No test processes"}
 const b=Math.max(1,Math.min(8,E[0]?.config?.poolSize??2)),S=E[0]?.config?.fingerprint||[]
 let C
-try{C=await(0,c.acquireSlot)(v,b)}catch(e){return{success:!1,message:"Could not acquire test worktree slot: "+(0,i.getErrorMessage)(e),error:(0,i.getErrorMessage)(e)}}const k=C.path,T=(0,n.simpleGit)(v)
+try{C=await(0,l.acquireSlot)(v,b)}catch(e){return{success:!1,message:"Could not acquire test worktree slot: "+(0,i.getErrorMessage)(e),error:(0,i.getErrorMessage)(e)}}const k=C.path,T=(0,n.simpleGit)(v)
 try{if((0,s.existsSync)(k))try{const e=(0,n.simpleGit)(k)
 await e.raw(["reset","--hard",I]),m.info("Reset pool slot to latest branch HEAD",{worktreePath:k,branchName:I,slotIndex:C.index})}catch(e){m.warn("Could not reset slot, recreating",{error:(0,i.getErrorMessage)(e)})
 try{await T.raw(["worktree","remove","--force",k])}catch{}}if(!(0,s.existsSync)(k)){(0,s.mkdirSync)((0,o.join)(v,".alchemy_lab","test-worktrees"),{recursive:!0})
-try{if(!(await T.branch()).all.includes(I))return(0,c.releaseSlot)(k),{success:!1,message:`Branch ${I} not found`,error:"Branch not found"}
-await T.raw(["worktree","add","--detach",k,I]),m.info("Created pool slot worktree (detached)",{worktreePath:k,branchName:I,slotIndex:C.index})}catch(e){if(!(e+"").includes("already exists"))return(0,c.releaseSlot)(k),{success:!1,message:"Failed to create pool slot worktree: "+(0,i.getErrorMessage)(e),error:(0,i.getErrorMessage)(e)}
-m.info("Pool slot worktree already exists, using as-is",{worktreePath:k})}}}catch(e){throw(0,c.releaseSlot)(k),e}const _=await(0,c.computeFingerprint)(k,S),j=0===S.length||_!==C.storedFingerprint,A=0===S.length?null:_===C.storedFingerprint
+try{if(!(await T.branch()).all.includes(I))return(0,l.releaseSlot)(k),{success:!1,message:`Branch ${I} not found`,error:"Branch not found"}
+await T.raw(["worktree","add","--detach",k,I]),m.info("Created pool slot worktree (detached)",{worktreePath:k,branchName:I,slotIndex:C.index})}catch(e){if(!(e+"").includes("already exists"))return(0,l.releaseSlot)(k),{success:!1,message:"Failed to create pool slot worktree: "+(0,i.getErrorMessage)(e),error:(0,i.getErrorMessage)(e)}
+m.info("Pool slot worktree already exists, using as-is",{worktreePath:k})}}}catch(e){throw(0,l.releaseSlot)(k),e}const _=await(0,l.computeFingerprint)(k,S),j=0===S.length||_!==C.storedFingerprint,A=0===S.length?null:_===C.storedFingerprint
 j||m.info("Pool slot fingerprint matches; skipping setup",{worktreePath:k,slotIndex:C.index,fingerprintFiles:S})
 const{execSync:P}=await Promise.resolve().then(t.t.bind(t,5317,23))
 if(j){for(const e of E){const t=e.config?.setup||[]
 if(0!==t.length){m.info("Running setup commands for test process",{processName:e.name,commandCount:t.length,slotIndex:C.index})
-for(const r of t)try{m.info("Setup command",{command:r,worktreePath:k}),P(r,{cwd:k,timeout:12e4,encoding:"utf-8",stdio:["pipe","pipe","pipe"],shell:"/bin/sh",env:{...process.env,FORCE_COLOR:"0"}})}catch(t){m.warn("Setup command failed",{command:r,processName:e.name,exitCode:t.status,stderr:(t.stderr||"").slice(-500)})}}}S.length>0&&await(0,c.writeSlotFingerprint)(k,_),await(0,c.writeWarmSnapshot)(k)}else{const{removed:e}=await(0,c.restoreToWarmSnapshot)(k)
+for(const r of t)try{m.info("Setup command",{command:r,worktreePath:k}),P(r,{cwd:k,timeout:12e4,encoding:"utf-8",stdio:["pipe","pipe","pipe"],shell:"/bin/sh",env:{...process.env,FORCE_COLOR:"0"}})}catch(t){m.warn("Setup command failed",{command:r,processName:e.name,exitCode:t.status,stderr:(t.stderr||"").slice(-500)})}}}S.length>0&&await(0,l.writeSlotFingerprint)(k,_),await(0,l.writeWarmSnapshot)(k)}else{const{removed:e}=await(0,l.restoreToWarmSnapshot)(k)
 e>0&&m.info("Pruned drift from warm slot",{worktreePath:k,slotIndex:C.index,removed:e})}let M
 try{const e=(0,n.simpleGit)(k)
 M=(await e.revparse(["HEAD"])).trim()}catch(e){m.warn("Failed to capture worktree HEAD; commit_sha will be NULL on this run",{worktreePath:k,error:(0,i.getErrorMessage)(e)})}const O=[],R=[],D=[]
@@ -1074,13 +1082,13 @@ for(const process of E){let t
 const s=e.testRunIds?.[O.length]
 try{s?(await f.patch(`/api/projects/${y}/test-runs/${s}`,$),t=(await f.get(`/api/projects/${y}/test-runs/${s}`)).data,O.push(t.id)):(t=(await f.post(`/api/projects/${y}/test-runs`,{requestId:d,testProcessId:process.id,parentSessionId:h,...$})).data,O.push(t.id))}catch(e){m.error("Failed to get/create test run record",{requestId:d,processId:process.id,preCreatedId:s,error:(0,i.getErrorMessage)(e)})
 continue}const o=r(g,process,0,y,p,e.commandNames,e.skipCommandNames),n=p?"tester":"auto-tester"
-try{const e=await(0,a.spawnActionAgent)({agent:n,prompt:o,cwd:k,projectPath:v,projectId:y,title:`[${d}] Test: ${process.name}`,requestId:d,parentSessionId:h,allowParallel:!0,apiClient:f,logger:m,trackChildPid:w})
+try{const e=await(0,c.spawnActionAgent)({agent:n,prompt:o,cwd:k,projectPath:v,projectId:y,title:`[${d}] Test: ${process.name}`,requestId:d,parentSessionId:h,allowParallel:!0,apiClient:f,logger:m,trackChildPid:w})
 if(!e.success)throw Error(e.error||"Spawn failed")
 R.push(t.id),e.sessionId&&D.push(e.sessionId),e.pid&&(x=e.pid),e.sessionId&&(f.patch(`/api/projects/${y}/test-runs/${t.id}`,{testerSessionId:e.sessionId}).catch((e=>m.warn("Failed to set testerSessionId on test run",{error:(0,i.getErrorMessage)(e)}))),f.post(`/api/projects/${y}/session-test-process-associations`,{opencodeSessionId:e.sessionId,testProcessId:process.id}).catch((t=>m.warn("Failed to create session\u2194test-process association",{sessionId:e.sessionId,testProcessId:process.id,error:(0,i.getErrorMessage)(t)})))),m.info("Spawned test-runner agent",{testRunId:t.id,processName:process.name,worktreePath:k,pid:e.pid,sessionId:e.sessionId})}catch(e){m.error("Failed to spawn test-runner agent",{testRunId:t.id,error:(0,i.getErrorMessage)(e)})
-try{await f.patch(`/api/projects/${y}/test-runs/${t.id}`,{status:"error",results:{steps:[],summary:"Failed to spawn test agent: "+(0,i.getErrorMessage)(e)}})}catch{}}}if(0===R.length)return(0,c.releaseSlot)(k),{success:!1,message:"Failed to spawn any test agents",error:"All agent spawns failed"}
-for(const e of D);return(0,c.watchPidAndRelease)(x,k,12e5),{success:!0,message:`Started ${R.length} test run(s) for request ${d}`,data:{testRunIds:O,agentsSpawned:R.length,processCount:E.length,worktreePath:k,spawnedSessionIds:D}}}catch(e){const t=(0,i.getErrorMessage)(e)
+try{await f.patch(`/api/projects/${y}/test-runs/${t.id}`,{status:"error",results:{steps:[],summary:"Failed to spawn test agent: "+(0,i.getErrorMessage)(e)}})}catch{}}}if(0===R.length)return(0,l.releaseSlot)(k),{success:!1,message:"Failed to spawn any test agents",error:"All agent spawns failed"}
+for(const e of D);return(0,l.watchPidAndRelease)(x,k,12e5),{success:!0,message:`Started ${R.length} test run(s) for request ${d}`,data:{testRunIds:O,agentsSpawned:R.length,processCount:E.length,worktreePath:k,spawnedSessionIds:D}}}catch(e){const t=(0,i.getErrorMessage)(e)
 return m.error("run-tests action failed",{requestId:d,error:t}),{success:!1,message:"Failed to run tests: "+t,error:t}}}
-const s=t(9896),o=t(6928),n=t(3807),i=t(8824),a=t(8148),c=t(489)},6563(e,exports,t){"use strict"
+const s=t(9896),o=t(6928),n=t(3807),i=t(8824),a=t(5454),c=t(8148),l=t(489)},6563(e,exports,t){"use strict"
 Object.defineProperty(exports,"__esModule",{value:!0}),exports.TimerHookScheduler=void 0
 const r=t(6175),s=t(4627),o=t(4769)
 class n{constructor(e){this.tracking=new Map,this.inFlight=new Set,this.paused=new Set,this.stopped=new Set,this.cachedHooks=[],this.lastHookFetch=0,this.lastActivityTime=0,this.agentTypeCache=new Map,this.logger=e.logger,this.hookExecutor=e.hookExecutor,this.apiClient=e.apiClient,this.getOpenCodeUrl=e.getOpenCodeUrl,this.getDefaultModel=e.getDefaultModel,this.projectId=e.projectId,this.projectPath=e.projectPath,this.conditionEvaluator=new r.TimerConditionEvaluator(e.logger)}start(){this.tickTimer||(this.logger.info("Timer hook scheduler starting",{tickInterval:1e4}),setTimeout((()=>this.tick().catch((()=>{}))),3e3),this.tickTimer=setInterval((()=>{this.tick().catch((()=>{}))}),1e4))}stop(){this.tickTimer&&(clearInterval(this.tickTimer),this.tickTimer=void 0),this.tracking.clear(),this.inFlight.clear(),this.paused.clear(),this.stopped.clear(),this.cachedHooks=[],this.logger.info("Timer hook scheduler stopped")}pauseHook(e){return!this.stopped.has(e)&&(this.paused.add(e),this.logger.info("Timer hook paused",{hookId:e}),!0)}resumeHook(e){const t=this.paused.delete(e)
@@ -1394,7 +1402,21 @@ return{model:{providerID:e,modelID:r.join("/")}}})():{}})})
 i.ok?n.info("Pushed notification to CLI session",{sessionId:f}):n.warn("Failed to push notification to CLI session",{status:i.status})}}catch(e){n.warn("Could not push notification to CLI session (may be offline)",{error:(0,r.getErrorMessage)(e)})}}catch(e){n.warn("Failed to notify target session, spawning work agent instead",{requestId:o,error:(0,r.getErrorMessage)(e)})}if(!p)try{const e=await(0,s.spawnActionAgent)({agent:"work",prompt:h,cwd:c,projectPath:c,projectId:a,title:`[${o}] ${d.title}`,requestId:o,parentSessionId:g,apiClient:i,logger:n,trackChildPid:l})
 m=e.success,m?n.info("Spawned work agent for request",{requestId:o,sessionId:e.sessionId}):n.warn("Failed to spawn work agent",{requestId:o,error:e.error})}catch(e){n.warn("Failed to spawn work agent",{requestId:o,error:(0,r.getErrorMessage)(e)})}return n.info("Approve-and-work-request action completed",{requestId:o,notifiedSession:p,agentSpawned:m}),{success:!0,message:`Successfully approved request ${o}: ${d.title}${p?" \u2014 creating session notified":m?" \u2014 work agent spawned":""}`,data:{operationType:"approve-and-work",previousStatus:"pending_approval",newStatus:"open",notifiedSession:p,agentSpawned:m}}}catch(e){const t=(0,r.getErrorMessage)(e)
 return n.error("approve-and-work-request action failed",{requestId:o,error:t}),{success:!1,message:"Failed to approve request: "+t,error:t}}}
-const r=t(8824),s=t(8148)},7904(e,exports){"use strict"
+const r=t(8824),s=t(8148)},7775(e,exports,t){"use strict"
+Object.defineProperty(exports,"__esModule",{value:!0}),exports.REVIEWER_FOCUSES=void 0,exports.handleSpawnReviewerAction=async function(e,t){const{requestId:n,parentSessionId:i,focus:a}=e,{logger:c,apiClient:l,projectId:d,projectPath:u,trackChildPid:h}=t
+if(!n||!i||!a)return{success:!1,message:"spawn-reviewer requires requestId, parentSessionId, and focus",error:"Missing payload fields"}
+if("string"!=typeof(g=a)||!exports.REVIEWER_FOCUSES.includes(g))return{success:!1,message:"Invalid focus: "+a,error:"Invalid focus"}
+var g
+c.info("Starting spawn-reviewer action",{requestId:n,focus:a,parentSessionId:i})
+let p=""
+if("CRITERIA"===a)try{const e=await l.get(`/api/projects/${d}/requests/${n}`)
+if(e.success&&e.data){const t=e.data,r=Array.isArray(t.acceptanceCriteria)?t.acceptanceCriteria:[],s=r.length>0?(0,o.formatCriteriaWithKeys)(r):"",n="string"==typeof t.completionSummary?t.completionSummary.trim():"",i=[]
+i.push(s?"Acceptance criteria to verify (shown as [key] [STATUS] text \u2014 record the ones you confirm satisfied via record_criteria_verification, citing the key):\n"+s:"This request has no acceptance criteria defined \u2014 report (none) for must_fix."),n&&i.push("Coder's completion summary (what they say they did):\n"+n),p=`\n${i.join("\n\n")}\n`}}catch(e){c.warn("spawn-reviewer: could not fetch request for CRITERIA context; proceeding without",{requestId:n,error:(0,s.getErrorMessage)(e)})}try{const e=function(e,t,r,s){const o=`[Session Context]\n- Request id: ${e}\n- Your focus: ${t}\n- Coder session to report to: ${r}`
+return"CRITERIA"===t?`${o}\n\nYou are the CRITERIA reviewer. You verify the diff actually delivers each acceptance criterion, RECORD your per-criterion verdict, then report. The other reviewers cover code quality; you cover whether the work is complete and correct against the spec.\n${s}\n1. Call \`request_diff\` to read the diff.\n2. Use \`read_files\` and \`search\` to confirm whether each criterion is satisfied by the change (cite file:line).\n3. For every criterion you confirm is satisfied, record it with \`record_criteria_verification\` { requestId: "${e}", verifications: [{ key, status, evidence }] } \u2014 record a verdict for EVERY criterion in a SINGLE call: status is "met" (evidence = file:line justification) or "unmet" (evidence = what is missing and how to fix it). Use the [key] shown next to each criterion.\n4. Any criterion you mark "unmet" is also a must_fix in your report. The server opens a rework task from the unmet criteria for the coder to address.\n5. Send ONE report via \`session_message_operations send\` with type "report", toSessionId "${r}", finish true, and a structured markdown summary (a per-criterion table + a must_fix section).\n6. Exit. Do not send follow-ups.`:`${o}\n\nYou are reviewing this request's diff with a single focus: **${t}**. Other dimensions are being reviewed in parallel by your peers \u2014 ignore them silently.\n\n1. Call \`request_diff\` to read the diff.\n2. Use \`read_files\` and \`search\` to gather context for any finding you intend to flag.\n3. Categorize each finding into must_fix / nice_to_have / out_of_scope per your agent instructions.\n4. Send ONE report via \`session_message_operations send\` with type "report", toSessionId "${r}", finish true, and the structured markdown summary.\n5. Exit. Do not send follow-ups.`}(n,a,i,p),t=`Reviewer (${a}) for ${n}`,s=await(0,r.spawnActionAgent)({agent:"CRITERIA"===a?"criteria-reviewer":"reviewer",prompt:e,cwd:u,projectPath:u,projectId:d,title:t,requestId:n,parentSessionId:i,allowParallel:!0,apiClient:l,logger:c,trackChildPid:h})
+return s.success&&s.sessionId?{success:!0,message:`Reviewer spawned: ${a} for ${n}`,data:{sessionId:s.sessionId,pid:s.pid,focus:a}}:{success:!1,message:`Failed to spawn reviewer (${a}): ${s.error}`,error:s.error||"Spawn failed"}}catch(e){const t=(0,s.getErrorMessage)(e)
+return c.error("spawn-reviewer action failed",{requestId:n,focus:a,error:t}),{success:!1,message:`Failed to spawn reviewer (${a}): ${t}`,error:t}}}
+const r=t(8148),s=t(8824),o=t(5454)
+exports.REVIEWER_FOCUSES=["REUSE","QUALITY","EFFICIENCY","CRITERIA"]},7904(e,exports){"use strict"
 Object.defineProperty(exports,"__esModule",{value:!0}),exports.REPORT_STATUS_EMOJI=void 0,exports.formatReportNotification=function(e){const{messageCode:t,status:r,summary:s,details:o,fromSessionId:n,senderAgent:i}=e,a=[`${exports.REPORT_STATUS_EMOJI[r]||"\u{1f4cb}"} **Report \u2014 ${r}**`]
 return t&&a.push("**Report Code:** "+t),i&&a.push(n?`from: ${i} (${n})`:"from: "+i),a.push("",s||"Task completed"),o&&a.push("",o),t&&a.push("",`Use \`session_message_operations\` with \`reply\` operation and message code \`${t}\` to acknowledge.`),a.join("\n")},exports.REPORT_STATUS_EMOJI={success:"\u2705",partial:"\u26a0\ufe0f",failure:"\u274c",blocked:"\u{1f6ab}"}},7960(e,exports){"use strict"
 function t(){const e=process.argv[1]||""
@@ -1582,7 +1604,7 @@ for(const e of T){if(!e.enabled)continue
 const t={}
 for(const[r,s]of Object.entries(e.tools||{})){const e=s,n=r.replace(/-/g,"_")
 t[o.OPENCODE_BUILTIN_TOOLS.has(r)?r:`${d}_${n}`]=e}const r=_?.metadata?.bashToolAgents||{}
-e.name in r&&(t.bash=!!r[e.name]),t.skill=!1,("subagent"===e.mode||"tester"===e.name||"auto-tester"===e.name)&&void 0===t.task&&(t.task=!1)
+e.name in r&&(t.bash=!!r[e.name]),t.skill=!1,("subagent"===e.mode||"tester"===e.name||"auto-tester"===e.name||"reviewer"===e.name||"criteria-reviewer"===e.name)&&void 0===t.task&&(t.task=!1)
 let s=""
 k&&(s+=k.trim()+"\n\n"),e.prompt&&(s+=e.prompt),e.promptAddition&&(s+="\n\n"+e.promptAddition),j.agent[e.name]={tools:t,...s?{prompt:s}:{},...e.description?{description:e.description}:{},mode:"utility"===e.mode?"primary":e.mode||"primary",...e.color?{color:e.color}:{}}}j.agent.general={...j.agent.general,disable:!0},j.agent.explore={...j.agent.explore,tools:{grep:!1,glob:!1,list:!1,read:!1,bash:!1,[d+"_search"]:!0,[d+"_read_files"]:!0,[d+"_view_asset"]:!0,[d+"_query_history"]:!0,webfetch:!0,websearch:!0,codesearch:!0}}
 const e=`---\ndescription: Fast codebase exploration and research subagent\nmode: subagent\n---\n\n${k?k.trim()+"\n\n":""}You are a codebase exploration agent. Your job is to find specific information and return concise, targeted answers.\n\n## Response Guidelines\n\n- **Be concise.** Return only what was asked for. A typical answer is 100-500 words.\n- **Lead with the answer.** State findings first, then supporting details.\n- **Reference specific files and lines.** Use \`file_path:line_number\` format.\n- **Don't exhaustively document everything you find.** Focus on what's relevant to the question.\n- **Use structured output** when listing multiple items (bullet points, not prose).\n\n## Strategy\n\n1. Start with \`search\` to find relevant files by name or content patterns\n2. Use \`read_files\` to read specific sections (use line ranges, don't read entire large files)\n3. Use \`query_history\` for understanding why code changed or tracing decisions\n\n## Anti-patterns to Avoid\n\n- Don't produce multi-thousand-word reports for simple questions\n- Don't read every file in a directory when a targeted search would suffice\n- Don't explain how tools work \u2014 just use them and report findings\n`