@tea-agent/loop-agent 0.44.0-next.12 → 0.44.0-next.13

package/dist/workflows/dag/hybrid/templates/frontend-test.js

@@ -5,6 +5,7 @@ import { assertValidDagSpec } from "../../validate.js";
 import { DEFAULT_DAG_EXECUTOR_MODELS, DEFAULT_DAG_OUTPUT_LANGUAGE, parseDagSpec } from "../../types.js";
 import { pathMatchesPattern } from "../../../../shared/git-progress.js";
 import { applyFrontendTestLayoutToText, resolveFrontendTestLayout } from "../../frontend-test-layout.js";
+import { isHarnessNamespacePath } from "../../test-artifact-namespace.js";
 import { buildFrontendTestOutcomeGateShellSnippet } from "../../frontend-test-result-contract.js";
 import { GENERATED_DAG_RUNTIME_CONTRACT, HYBRID_DEFAULTS, STANDARD_GLOBAL_CONSTRAINTS, buildDagSourceBinding, buildSourceContextBlock, extractObjective, extractSuccessCriteria, resolveDagVerifyStrategy } from "../sources.js";
 import { applyDefaultReadOnlyRetryPolicy, commonForbiddenPaths, commonReadOnlyPaths, stampGeneratedArtifactBindings, } from "./shared.js";
@@ -62,7 +63,7 @@ export function buildFrontendTestHybridDag(sources) {
     const maxCasesPerBatchLiteral = String(config.maxCasesPerBatch);
     const checklistScript = [
         "const fs=require('fs'),path=require('path');",
-        "const root='testcase/frontend/cases';",
+        `const root='${layout.casesDir}';`,
         "const draft=path.join(root,'manifest.draft.json');",
         "const final=path.join(root,'manifest.json');",
         "const manifestPath=fs.existsSync(draft)?draft:(fs.existsSync(final)?final:null);",
@@ -84,8 +85,8 @@ export function buildFrontendTestHybridDag(sources) {
         " if(typeof c.caseId==='string'&&/^AC-/i.test(c.caseId))issues.push({ruleId:'case-id-is-ac',caseId:id,detail:'do not use acceptance id as caseId; put AC-FE-* only in acIds'});",
         " const casePath=typeof c.casePath==='string'?c.casePath:null;",
         " if(!casePath||!fs.existsSync(casePath)){issues.push({ruleId:'case-file-missing',caseId:id,detail:String(casePath)});continue;}",
-        " if(typeof c.caseId==='string'&&casePath!=='testcase/frontend/cases/'+c.caseId+'.md')issues.push({ruleId:'case-path-mismatch',caseId:id,detail:casePath+' must equal testcase/frontend/cases/'+c.caseId+'.md'});",
-        " const body=fs.readFileSync(casePath,'utf8');",
+        " if(typeof c.caseId==='string'&&casePath!=='${layout.casesDir}/'+c.caseId+'.md')issues.push({ruleId:'case-path-mismatch',caseId:id,detail:casePath+' must equal ${layout.casesDir}/'+c.caseId+'.md'});",
+        " const body=fs.readFileSync(casePath,'utf8').replace(/\\bplaywright_cli\\b/g,'playwright-cli');",
         " if(!openRe.test(body))issues.push({ruleId:'open-prefix',caseId:id,detail:'missing playwright-cli open --browser=chrome <absolute-url>; playwright-cli is strongly recommended for browser execution'});",
         ,
         " const m=body.match(/playwright-cli\\s+open\\s+--browser=chrome\\s+(https?:\\/\\/\\S+)/i);",
@@ -117,7 +118,7 @@ export function buildFrontendTestHybridDag(sources) {
         JSON.stringify([
             "const fs=require('fs'),path=require('path');",
             "function fail(ruleId,detail){ const msg=JSON.stringify({ruleId:ruleId,detail:String(detail||'')}); console.error('frontend-test manifest blocked: '+msg); throw new Error('frontend-test manifest blocked: '+ruleId+(detail?(': '+detail):'')); }",
-            "const draft='testcase/frontend/cases/manifest.draft.json',file='testcase/frontend/cases/manifest.json',tmp=file+'.tmp';",
+            `const draft='${layout.casesDir}/manifest.draft.json',file='${layout.casesDir}/manifest.json',tmp=file+'.tmp';`,
             "if(!fs.existsSync(draft)) fail('draft-missing','missing '+draft);",
             "let manifest; try{manifest=JSON.parse(fs.readFileSync(draft,'utf8'));}catch(e){fail('draft-invalid-json',e&&e.message||e);}",
             "if(manifest.schemaVersion!==1) fail('draft-schema','schemaVersion must be 1');",
@@ -138,8 +139,8 @@ export function buildFrontendTestHybridDag(sources) {
             " if(typeof c.dimension!=='string'||!dims.has(c.dimension)) fail('invalid-dimension',String(c.dimension));",
             " if(!Array.isArray(c.acIds)||c.acIds.length===0||c.acIds.some(a=>typeof a!=='string'||!a.trim())) fail('ac-mapping','invalid acIds for '+c.caseId);",
             " for(const ac of c.acIds){ if(!acIdRe.test(ac)) fail('ac-id-shape','acIds entry must be AC-* acceptance id, not caseId: '+ac); if(truncatedAcRe.test(ac)||(declaredAc.size===0&&truncatedUndeclaredAcRe.test(ac))) fail('ac-id-truncated','acIds entry is a truncated acceptance id (missing feature prefix or sequence number): '+ac); if(declaredAc.size>0&&!declaredAc.has(ac)) fail('unknown-ac',ac+' not in sourceBinding; repair generator input or AC list'); }",
-            " c.casePath='testcase/frontend/cases/'+c.caseId+'.md';",
-            " c.evidenceDir='testcase/frontend/evidence/'+c.caseId+'/';",
+            ` c.casePath='${layout.casesDir}/'+c.caseId+'.md';`,
+            ` c.evidenceDir='${layout.evidenceDir}/'+c.caseId+'/';`,
             " for(const k of ['casePath','evidenceDir']){ const v=c[k]; if(typeof v!=='string'||path.isAbsolute(v)||v.includes('..')) fail('unsafe-path',k+': '+String(v)); }",
             " if(!fs.existsSync(c.casePath)) fail('case-file-missing','missing case file '+c.casePath+' (filename must equal caseId.md)');",
             " if(seenCasePath.has(c.casePath)) fail('duplicate-case-path',c.casePath); seenCasePath.add(c.casePath);",
@@ -158,7 +159,7 @@ export function buildFrontendTestHybridDag(sources) {
             "const runDir=process.env.HARNESS_DAG_RUN_DIR; if(!runDir)throw new Error('missing HARNESS_DAG_RUN_DIR');",
             "const resultPath=path.join(runDir,'contracts','frontend-test-result.json'); if(!fs.existsSync(resultPath))throw new Error('missing '+resultPath);",
             "const r=JSON.parse(fs.readFileSync(resultPath,'utf8')); const findings=Array.isArray(r.advisoryFindings)?r.advisoryFindings:[];",
-            "const reports='testcase/frontend/reports'; fs.mkdirSync(reports,{recursive:true});",
+            `const reports='${layout.reportsDir}'; fs.mkdirSync(reports,{recursive:true});`,
             "const esc=v=>String(v??'').replaceAll('&','&amp;').replaceAll('<','&lt;').replaceAll('>','&gt;').replaceAll('\\\"','&quot;');",
             "const labels={cases:'用例总数',passed:'通过',failed:'失败',blocked:'阻塞'}; const totals=r.totals||{}; const missing=r.acceptanceCoverage&&Array.isArray(r.acceptanceCoverage.missing)?r.acceptanceCoverage.missing:[];",
             "const md=['# 前端测试质量建议报告','','> Advisory：以下 finding 用于改进测试资产和证据质量，不阻塞后续流程。','','## 执行摘要','',...Object.entries(labels).map(([k,v])=>'- '+v+': '+Number(totals[k]||0)),'- 执行结果: '+String(r.outcome||'unknown'),'- 缺失 AC: '+(missing.join(', ')||'无'),'','## 建议项','',...(findings.length?findings.map(f=>'- ['+f.ruleId+']'+(f.caseId?' '+f.caseId:'')+': '+f.detail):['- 未发现建议项。']),''];",
@@ -197,7 +198,7 @@ export function buildFrontendTestHybridDag(sources) {
             writeSet: ragWriteSet,
             allowedPaths: [...ragWriteSet],
             forbiddenPaths: forbidden,
-            outputContract: "Write testcase/frontend/rag/standard-scenarios.v1.json for generate-time standard scenario coverage. Copy docs/templates or harness.json governanceRoot templates (including ai_workspace/loop-agent/templates) when present; otherwise write the minimal STD-FE-SMOKE-ENTRY fallback.",
+            outputContract: `Write ${layout.ragDir}/standard-scenarios.v1.json for generate-time standard scenario coverage. Copy docs/templates or harness.json governanceRoot templates (including ai_workspace/loop-agent/templates) when present; otherwise write the minimal STD-FE-SMOKE-ENTRY fallback.`,
             subtask_prompt: "Prepare frontend-test package: materialize standard-scenarios.v1.json into the RAG package from docs/templates, governanceRoot/templates, or the init-projected ai_workspace/loop-agent/templates path.",
             shell: {
                 commands: [],
@@ -218,10 +219,10 @@ export function buildFrontendTestHybridDag(sources) {
             writeSet: ragWriteSet,
             allowedPaths: [...commonReadOnlyPaths(sources), ...ragWriteSet],
             forbiddenPaths: forbidden,
-            outputContract: "Write short testcase/frontend/rag/context.md and coverage-map.md with a model-resolved non-production baseUrl/baseUrlSource from task source config, environmentProbe=pending, and capability notes.",
+            outputContract: `Write short ${layout.ragDir}/context.md and coverage-map.md with a model-resolved non-production baseUrl/baseUrlSource from task source config, environmentProbe=pending, and capability notes.`,
             subtask_prompt: [
                 "Build the frontend test RAG package (keep it short).",
-                "Read task source, routes/components/API or Mock facts, and execution contract. Write only testcase/frontend/rag/context.md and coverage-map.md.",
+                `Read task source, routes/components/API or Mock facts, and execution contract. Write only ${layout.ragDir}/context.md and coverage-map.md.`,
                 "Prefer fixed fields: baseUrl, baseUrlSource, environmentProbe, AC table, capability matrix (backend real/mock, pagination data, HTTP observation, error injection), risks, forbidden hosts. Do not paste large implementation dumps.",
                 "Base URL resolution (required): read the bound task source and config/environment references. Prefer TARGET_URL/targetUrl as the concrete tested entry, then BASE_URL/baseUrl or FRONTEND_BASE_URL/frontendUrl, then LOGIN_URL/loginUrl; accept common casing and underscore/kebab/space variants. Choose one absolute non-production http(s) URL, never API-only or production hosts. If no usable URL exists, use http://localhost:5173. Write `baseUrl: <url>`, `baseUrlSource: <bound source path>|default-localhost-5173`, and `environmentProbe: pending`. Include exact start prefix: playwright-cli open --browser=chrome <resolved-base-url> with the concrete selected URL.",
                 buildSourceContextBlock(sources),
@@ -238,7 +239,7 @@ export function buildFrontendTestHybridDag(sources) {
             allowedPaths: [...ragWriteSet],
             forbiddenPaths: forbidden,
             outputContract: "Fail-closed environment preflight: absolute non-production baseUrl + curl HTTP reachability; writes environmentProbe facts; unreachable => blockedReason frontend-base-url-unreachable with errorClass (connection-refused / dns-unresolved / connect-timeout / http-N / curl-exit-N). Node ERROR so generate/map do not run. Does not start the app.",
-            subtask_prompt: "Parse the resolved absolute baseUrl from testcase/frontend/rag/context.md, reject production/non-http(s)/credential/query/fragment URLs, then probe it with curl HEAD and GET fallback (connect/max-time; no auth/cookie). 2xx/3xx => reachable and continue. Connection refused (curl 7) records errorClass=connection-refused and tells the operator to start the local app (scripts/serve.sh or npm start) then rerun from this node. 4xx/5xx/DNS/timeout/TLS => blockedReason frontend-base-url-unreachable with a distinct errorClass. Missing curl => blockedReason curl-unavailable. Do not start the app. Fixture/reset remain soft guidance.",
+            subtask_prompt: `Parse the resolved absolute baseUrl from ${layout.ragDir}/context.md, reject production/non-http(s)/credential/query/fragment URLs, then probe it with curl HEAD and GET fallback (connect/max-time; no auth/cookie). 2xx/3xx => reachable and continue. Connection refused (curl 7) records errorClass=connection-refused and tells the operator to start the local app (scripts/serve.sh or npm start) then rerun from this node. 4xx/5xx/DNS/timeout/TLS => blockedReason frontend-base-url-unreachable with a distinct errorClass. Missing curl => blockedReason curl-unavailable. Do not start the app. Fixture/reset remain soft guidance.`,
             shell: {
                 commands: [],
                 frontendTestEnvironmentProbe: {},
@@ -261,22 +262,22 @@ export function buildFrontendTestHybridDag(sources) {
             outputContract: "Write executable Markdown frontend cases, index.md, and manifest.draft.json schemaVersion 1 only; the exclusive shell materializer promotes the validated draft to manifest.json. Do not write manifest.json or test source code.",
             subtask_prompt: [
                 "Use skill playwright-cli-case-generator.",
-                "Read testcase/frontend/rag/standard-scenarios.v1.json and cover priority=must scenarios (or record GAP in coverage-map). Include ## 测试点 and ## 测试步骤 in each case.",
+                `Read ${layout.ragDir}/standard-scenarios.v1.json and cover priority=must scenarios (or record GAP in coverage-map). Include ## 测试点 and ## 测试步骤 in each case.`,
                 ...(uiAnchorsLedgerInstruction ? [uiAnchorsLedgerInstruction] : []),
-                "Read only testcase/frontend/rag/context.md, testcase/frontend/rag/coverage-map.md, and the draft case paths testcase/frontend/cases/FE-*.md, testcase/frontend/cases/index.md, and testcase/frontend/cases/manifest.draft.json. Write only those same draft paths. Do not write testcase/frontend/cases/manifest.json.",
+                `Read only ${layout.ragDir}/context.md, ${layout.ragDir}/coverage-map.md, and the draft case paths ${layout.casesDir}/FE-*.md, ${layout.casesDir}/index.md, and ${layout.casesDir}/manifest.draft.json. Write only those same draft paths. Do not write ${layout.casesDir}/manifest.json.`,
                 "Generate Markdown cases, index.md and manifest.draft.json (schemaVersion 1; cases[] with caseId, casePath, dimension, acIds, evidenceDir).",
                 "HARD ID CONTRACT (do not confuse these):",
                 "- caseId / filename MUST be FE-<FEATURE>-<NNN>-<dimension> (example FE-LOGIN-001-core). NEVER use AC-FE-* as caseId or filename.",
                 "- acIds MUST list acceptance criteria only: AC-FE-* / AC-* from the declared task list (example AC-FE-001). NEVER put FE-* case ids into acIds.",
-                "- casePath MUST equal testcase/frontend/cases/<caseId>.md; evidenceDir MUST equal testcase/frontend/evidence/<caseId>/. Materialize will rewrite paths, but files must already use caseId filenames.",
+                `- casePath MUST equal ${layout.casesDir}/<caseId>.md; evidenceDir MUST equal ${layout.evidenceDir}/<caseId>/. Materialize will rewrite paths, but files must already use caseId filenames.`,
                 `Declared acceptance ids for this task (use only these in acIds when non-empty): ${declaredAcIds.length > 0 ? declaredAcIds.join(", ") : "(none extracted - still use AC-* shape, never FE-* case ids)"}.`,
                 "dimensions: core|boundary|flow|backend only.",
                 "Prefer a small smoke suite (default max roughly 4-8 cases unless task frontendTest.maxCasesPerBatch is higher). Never invent unavailable API fields or credentials. Do not create pytest or Playwright source.",
                 "HARD playwright-cli-only: every browser step must use repo skill playwright-cli declared commands only. Forbidden: bare `playwright`, `npx playwright`, `playwright test`, `@playwright/test`, Node Playwright API, or generating Playwright/Pytest source. No fallback when playwright-cli is unavailable - case must instruct blocked evidence playwright-cli-unavailable.",
-                "Copy the resolved absolute baseUrl from testcase/frontend/rag/context.md after the environment probe has marked it reachable. Every browser start command must be exactly: playwright-cli open --browser=chrome <resolved-base-url-from-context.md> with that concrete URL. Never leave a base-url placeholder. Use default browser session only; never write -s=<case-id>.",
+                `Copy the resolved absolute baseUrl from ${layout.ragDir}/context.md after the environment probe has marked it reachable. Every browser start command must be exactly: playwright-cli open --browser=chrome <resolved-base-url-from-context.md> with that concrete URL. Never leave a base-url placeholder. Use default browser session only; never write -s=<case-id>.`,
                 "HARD dynamic refs: executable playwright-cli lines must never contain an angle-bracket token such as <fresh-ref> or descriptive <...> placeholder. Use only shell-safe documentation placeholders `eX`, `eY`, ...; each means the real `eNN` ref parsed from the immediately preceding latest `snapshot`. Write a fresh snapshot before every element reference. eX/eY are never literal structured-tool arguments; a later snapshot invalidates prior refs, so never reuse stale refs.",
                 "HARD file-output argv: use canonical `--filename` only. Screenshot uses `playwright-cli screenshot --filename final.png` (a real ref may precede the flag); PDF uses `playwright-cli pdf --filename final.pdf`; snapshot without filename is response-only and a snapshot file uses `playwright-cli snapshot --filename snapshot.txt`. Never generate `playwright-cli screenshot <path>`, use `--path`, `--output`, or `--file`, or pass an output path as a positional target.",
-                "Each case must be independently reproducible with fixture/reset, UI reset, snapshot-before-ref, evidence write point under testcase/frontend/evidence/<case-id>/. If the isolated environment is unavailable, require writing blocked evidence before any browser command.",
+                `Each case must be independently reproducible with fixture/reset, UI reset, snapshot-before-ref, evidence write point under ${layout.evidenceDir}/<case-id>/. If the isolated environment is unavailable, require writing blocked evidence before any browser command.`,
                 "U/D data ownership (required for modify/delete): (1) Only mutate data whose ownership is proven by current login identity + observable UI/API owner fields - never by name/guessed id/list order alone. (2) If current user has no data, create tagged cleanable data in current-user context, then U/D, then cleanup+verify. (3) Else only task-authorized Mock, labeled as Mock (not real backend proof). (4) If ownership unverifiable and create/Mock unavailable: write blocked with blockedReason current-user-data-unavailable | data-ownership-unverifiable | safe-test-data-setup-unavailable - do not risk cross-user data. (5) Never touch other users, shared fixtures, production, or non-cleanable data.",
             ].join("\n\n"),
         },
@@ -307,7 +308,7 @@ export function buildFrontendTestHybridDag(sources) {
             allowedPaths: [...ragWriteSet, ...caseDraftWriteSet],
             forbiddenPaths: forbidden,
             outputContract: "Apply the one permitted frontend case revision to FE-*.md, index.md, and manifest.draft.json only; the exclusive shell materializer remains the sole writer of manifest.json. No browser execution or evidence writes.",
-            subtask_prompt: "This is the only permitted case revision. Read the first review findings and the RAG package. Revise only testcase/frontend/cases/FE-*.md, testcase/frontend/cases/index.md, and testcase/frontend/cases/manifest.draft.json; preserve traceable AC mappings. Preserve the dynamic-ref contract: executable playwright-cli lines use only eX/eY-style shell-safe documentation placeholders, never <...>; each placeholder is resolved from the immediately preceding latest snapshot and stale refs are not reused. Do not write testcase/frontend/cases/manifest.json, execute a browser, or write evidence. HARD: Never delete case files; only edit in place or add missing cases. Preserve the full planned suite, index.md, and manifest.draft.json.",
+            subtask_prompt: `This is the only permitted case revision. Read the first review findings and the RAG package. Revise only ${layout.casesDir}/FE-*.md, ${layout.casesDir}/index.md, and ${layout.casesDir}/manifest.draft.json; preserve traceable AC mappings. Preserve the dynamic-ref contract: executable playwright-cli lines use only eX/eY-style shell-safe documentation placeholders, never <...>; each placeholder is resolved from the immediately preceding latest snapshot and stale refs are not reused. Do not write ${layout.casesDir}/manifest.json, execute a browser, or write evidence. HARD: Never delete case files; only edit in place or add missing cases. Preserve the full planned suite, index.md, and manifest.draft.json.`,
         }, {
             id: "review-frontend-cases-final-pi",
             depends_on: ["revise-frontend-cases-pi"],
@@ -411,9 +412,9 @@ export function buildFrontendTestHybridDag(sources) {
                 writePolicy: "exclusive",
                 writeGuardPolicy: "tools-only",
                 allowedPaths: [
-                    "testcase/frontend/cases/{{case.caseId}}.md",
-                    "testcase/frontend/rag/context.md",
-                    "testcase/frontend/rag/coverage-map.md",
+                    `${layout.casesDir}/{{case.caseId}}.md`,
+                    `${layout.ragDir}/context.md`,
+                    `${layout.ragDir}/coverage-map.md`,
                     `${evidenceRoot}/{{case.caseId}}/**`,
                 ],
                 forbiddenPaths: forbidden,
@@ -422,7 +423,7 @@ export function buildFrontendTestHybridDag(sources) {
                 subtaskPromptTemplate: [
                     "Primary job: EXECUTE case {{case.caseId}} from {{case.casePath}} with skill playwright-cli (fresh Pi session; do not use /new). playwright-cli-only: never bare Playwright CLI/API/test runner; no fallback.",
                     "Use the structured playwright_cli tool for every browser action. Do not request or search for bash. Do not execute raw shell commands. Translate each playwright-cli line in the case Markdown into one playwright_cli tool call ({command, args?, timeoutSeconds?}).",
-                    "1) Read the concrete baseUrl from testcase/frontend/rag/context.md; it has already passed the environment shell safety/reachability gate. Start browser ONLY via playwright_cli command=open with args [--browser=chrome, <that-concrete-baseUrl>] (default session only; no -s=). 2) Dynamic refs: eX/eY in case Markdown are documentation placeholders, never tool args. Immediately before every structured playwright_cli call that references an element, parse the current real eNN from the immediately preceding latest snapshot and pass only that real eNN; never send literal `eX`/`eY`. A new snapshot invalidates prior refs, so never reuse stale refs. File outputs are canonical: screenshot args [--filename, final.png] (or [e5, --filename, final.png] for a real target), PDF args [--filename, final.pdf], and snapshot writes a file only with [--filename, snapshot.txt]; snapshot without filename is response-only. Never use --path, --output, --file, or any output path as a positional target. Follow case steps with snapshot before element refs using only playwright_cli. A passed case requires this same child receipt order: successful open → successful find → controller post-execution cleanup. Only successful find is a meaningful assertion; snapshot, goto, screenshot, request/console, click/fill and other ordinary interactions cannot establish passed authority. 3) Only when preflight or playwright_cli tool explicitly fails may you write blocked evidence (blockedReason playwright-cli-unavailable | frontend-base-url-unreachable); never invent CLI-unavailable solely because bash is absent. Unique-control early stop: when an AC names a specific control that is absent from the first post-navigation snapshot after reaching the required state (and no setup-required gate appeared), do ONE find with the case's literal as evidence; if it returns 0 matches, record status=blocked with blockedReason unique-control-unavailable citing that single find - do NOT re-explore with alternative literals, repeated snapshots, navigation detours, or rerun loops. 4) For U/D: enforce current-user ownership / create-or-mock-or-blocked; never mutate other users' data.",
+                    `1) Read the concrete baseUrl from ${layout.ragDir}/context.md; it has already passed the environment shell safety/reachability gate. Start browser ONLY via playwright_cli command=open with args [--browser=chrome, <that-concrete-baseUrl>] (default session only; no -s=). 2) Dynamic refs: eX/eY in case Markdown are documentation placeholders, never tool args. Immediately before every structured playwright_cli call that references an element, parse the current real eNN from the immediately preceding latest snapshot and pass only that real eNN; never send literal \`eX\`/\`eY\`. A new snapshot invalidates prior refs, so never reuse stale refs. File outputs are canonical: screenshot args [--filename, final.png] (or [e5, --filename, final.png] for a real target), PDF args [--filename, final.pdf], and snapshot writes a file only with [--filename, snapshot.txt]; snapshot without filename is response-only. Never use --path, --output, --file, or any output path as a positional target. Follow case steps with snapshot before element refs using only playwright_cli. A passed case requires this same child receipt order: successful open → successful find → controller post-execution cleanup. Only successful find is a meaningful assertion; snapshot, goto, screenshot, request/console, click/fill and other ordinary interactions cannot establish passed authority. 3) Only when preflight or playwright_cli tool explicitly fails may you write blocked evidence (blockedReason playwright-cli-unavailable | frontend-base-url-unreachable); never invent CLI-unavailable solely because bash is absent. Unique-control early stop: when an AC names a specific control that is absent from the first post-navigation snapshot after reaching the required state (and no setup-required gate appeared), do ONE find with the case's literal as evidence; if it returns 0 matches, record status=blocked with blockedReason unique-control-unavailable citing that single find - do NOT re-explore with alternative literals, repeated snapshots, navigation detours, or rerun loops. 4) For U/D: enforce current-user ownership / create-or-mock-or-blocked; never mutate other users' data.`,
                     "Always write {{case.evidenceDir}}execution.md and {{case.evidenceDir}}case-result.json with caseId={{case.caseId}}, status passed|failed|blocked, evidencePaths (relative under evidenceDir). blocked needs non-empty blockedReason. After writing, self-check the same contract; if self-check fails, rewrite both files as status=blocked blockedReason=invalid-evidence-shape (never leave missing/malformed evidence).",
                     "Business failed/blocked is a recorded result, not a node failure. Close browser via playwright_cli command=close. Return compact JSON (<=1200 chars): {caseId,status,evidencePaths,errorSummary,tokens}.",
                 ].join("\n\n"),
@@ -442,11 +443,11 @@ export function buildFrontendTestHybridDag(sources) {
             : `rerun-candidates-round${round}.json`;
         const selectCommand = [
             "const fs=require('fs'),path=require('path');",
-            "const manifestPath='testcase/frontend/cases/manifest.json';",
+            `const manifestPath='${layout.casesDir}/manifest.json';`,
             "if(!fs.existsSync(manifestPath)){process.stdout.write(JSON.stringify({cases:[]}));process.exit(0);}",
             "const manifest=JSON.parse(fs.readFileSync(manifestPath,'utf8'));const cases=[];",
-            "for(const c of (manifest.cases||[])){const evidenceDir=(c.evidenceDir||('testcase/frontend/evidence/'+c.caseId+'/')).replace(/\\/+$/,'')+'/';const resultPath=path.join(evidenceDir,'case-result.json');const execPath=path.join(evidenceDir,'execution.md');let reason=null;let attempt=0;let missing=false;if(!fs.existsSync(resultPath)){missing=true;reason='missing-result-files';}else{try{const r=JSON.parse(fs.readFileSync(resultPath,'utf8'));attempt=Number(r.rerunAttempt||0)||0;if(r.status==='blocked')reason='blocked';if(r.status==='failed')reason='failed-retry';if(!r.status){missing=true;reason='missing-result-files';}}catch(_){missing=true;reason='missing-result-files';}}if(!fs.existsSync(execPath)&&reason!=='blocked'){missing=true;reason=reason||'missing-result-files';}const should=(reason==='blocked'||reason==='failed-retry'||missing)&&attempt<" + String(round) + ";if(should){cases.push({caseId:c.caseId,casePath:c.casePath||('testcase/frontend/cases/'+c.caseId+'.md'),evidenceDir,dimension:c.dimension||'core',acIds:c.acIds||[],rerunAttempt:attempt+1,reason:reason||'blocked'});}}",
-            `fs.mkdirSync('testcase/frontend/evidence',{recursive:true});fs.writeFileSync('testcase/frontend/evidence/${candidateArtifact}',JSON.stringify({schemaVersion:1,cases},null,2)+'\\n');process.stdout.write(JSON.stringify({cases}));`,
+            `for(const c of (manifest.cases||[])){const evidenceDir=(c.evidenceDir||('${layout.evidenceDir}/'+c.caseId+'/')).replace(/\\/+$/,'')+'/';const resultPath=path.join(evidenceDir,'case-result.json');const execPath=path.join(evidenceDir,'execution.md');let reason=null;let attempt=0;let missing=false;if(!fs.existsSync(resultPath)){missing=true;reason='missing-result-files';}else{try{const r=JSON.parse(fs.readFileSync(resultPath,'utf8'));attempt=Number(r.rerunAttempt||0)||0;if(r.status==='blocked')reason='blocked';if(r.status==='failed')reason='failed-retry';if(!r.status){missing=true;reason='missing-result-files';}}catch(_){missing=true;reason='missing-result-files';}}if(!fs.existsSync(execPath)&&reason!=='blocked'){missing=true;reason=reason||'missing-result-files';}const should=(reason==='blocked'||reason==='failed-retry'||missing)&&attempt<` + String(round) + `;if(should){cases.push({caseId:c.caseId,casePath:c.casePath||('${layout.casesDir}/'+c.caseId+'.md'),evidenceDir,dimension:c.dimension||'core',acIds:c.acIds||[],rerunAttempt:attempt+1,reason:reason||'blocked'});}}`,
+            `fs.mkdirSync('${layout.evidenceDir}',{recursive:true});fs.writeFileSync('${layout.evidenceDir}/${candidateArtifact}',JSON.stringify({schemaVersion:1,cases},null,2)+'\\n');process.stdout.write(JSON.stringify({cases}));`,
         ].join("");
         tasks.push({
             id: selectorId,
@@ -455,8 +456,8 @@ export function buildFrontendTestHybridDag(sources) {
             executor: "shell",
             complexity: "LOW",
             writePolicy: "exclusive",
-            writeSet: ["testcase/frontend/evidence/**"],
-            allowedPaths: [...casesWriteSet, "testcase/frontend/evidence/**"],
+            writeSet: [`${evidenceRoot}/**`],
+            allowedPaths: [...casesWriteSet, `${evidenceRoot}/**`],
             forbiddenPaths: forbidden,
             outputContract: `Stdout JSON {cases} for blocked or missing-result-file cases eligible for rerun round ${round}/${maxRerunAttempts}.`,
             subtask_prompt: "Select frontend-test cases eligible for this bounded rerun round.",
@@ -501,9 +502,9 @@ export function buildFrontendTestHybridDag(sources) {
                     writePolicy: "exclusive",
                     writeGuardPolicy: "tools-only",
                     allowedPaths: [
-                        "testcase/frontend/cases/{{case.caseId}}.md",
-                        "testcase/frontend/rag/context.md",
-                        "testcase/frontend/rag/coverage-map.md",
+                        `${layout.casesDir}/{{case.caseId}}.md`,
+                        `${layout.ragDir}/context.md`,
+                        `${layout.ragDir}/coverage-map.md`,
                         `${evidenceRoot}/{{case.caseId}}/**`,
                     ],
                     forbiddenPaths: forbidden,
@@ -512,7 +513,7 @@ export function buildFrontendTestHybridDag(sources) {
                     subtaskPromptTemplate: [
                         "RERUN attempt {{case.rerunAttempt}} for {{case.caseId}} (reason={{case.reason}}). Rewrite the authoritative execution.md and case-result.json; its final status replaces the earlier case result.",
                         "Primary job: EXECUTE case {{case.caseId}} from {{case.casePath}} with skill playwright-cli (fresh Pi session). Use structured playwright_cli only; headless open.",
-                        "Read the concrete baseUrl from testcase/frontend/rag/context.md and start via playwright_cli command=open with args [--browser=chrome, <that-concrete-baseUrl>]. Passed requires open → find → cleanup receipts.",
+                        `Read the concrete baseUrl from ${layout.ragDir}/context.md and start via playwright_cli command=open with args [--browser=chrome, <that-concrete-baseUrl>]. Passed requires open → find → cleanup receipts.`,
                         "Unique-control early stop: if the first post-navigation snapshot lacks the control an AC names (and no setup-required gate appeared), one find returning 0 matches is sufficient evidence - write status=blocked blockedReason unique-control-unavailable instead of exploratory retries.",
                         "Always write {{case.evidenceDir}}execution.md and {{case.evidenceDir}}case-result.json with caseId, status, evidencePaths, rerunAttempt={{case.rerunAttempt}}. Write fixed sections `### 执行摘要` and `### 实际执行步骤` to execution.md when available.",
                     ].join("\n\n"),
@@ -531,7 +532,7 @@ export function buildFrontendTestHybridDag(sources) {
         complexity: "LOW",
         writePolicy: "exclusive",
         writeSet: [`${evidenceRoot}/**`],
-        allowedPaths: ["testcase/frontend/cases/**", `${evidenceRoot}/**`],
+        allowedPaths: [`${layout.casesDir}/**`, `${evidenceRoot}/**`],
         forbiddenPaths: forbidden,
         outputContract: "Evidence validation (advisory missing/malformed does not fail the node; hard-fail only path escape) then hash-bound frontend-test-result-v1. Node success means result-v1 was written, not that all cases passed.",
         subtask_prompt: "Validate case evidence then materialize authoritative frontend-test-result-v1. Missing/malformed case evidence is advisory; only unsafe evidence paths hard-fail. Do not use Pi prose as input.",
@@ -549,8 +550,8 @@ export function buildFrontendTestHybridDag(sources) {
         executor: "shell",
         complexity: "LOW",
         writePolicy: "exclusive",
-        writeSet: ["testcase/frontend/reports/**"],
-        allowedPaths: ["testcase/frontend/**"],
+        writeSet: [`${layout.reportsDir}/**`],
+        allowedPaths: [`${layout.testRoot}/**`],
         forbiddenPaths: forbidden,
         outputContract: "Deterministic L-5 (optional) plus main frontend-test-report.md/html from frontend-test-result-v1. Pipeline acceptance = result-v1 + main HTML.",
         subtask_prompt: "Render operational reports from frontend-test-result-v1 only. Do not invent coverage. Main HTML is required; L-5 follows frontendTest.reports.l5.",
@@ -596,10 +597,10 @@ export function buildFrontendTestHybridDag(sources) {
             complexity: "MED",
             writePolicy: "exclusive",
             writeGuardPolicy: "tools-only",
-            writeSet: ["testcase/frontend/reports/**"],
-            allowedPaths: ["testcase/frontend/**"],
+            writeSet: [`${layout.reportsDir}/**`],
+            allowedPaths: [`${layout.testRoot}/**`],
             forbiddenPaths: forbidden,
-            outputContract: "Optional retrospective under testcase/frontend/reports/frontend-test-retrospect-<date>.md. Not required for pipeline acceptance.",
+            outputContract: `Optional retrospective under ${layout.reportsDir}/frontend-test-retrospect-<date>.md. Not required for pipeline acceptance.`,
             subtask_prompt: "Write optional frontend-test retrospective after result + reports. Do not recompute L-5. Pipeline success does not require this file.",
         });
     }
@@ -617,7 +618,7 @@ export function buildFrontendTestHybridDag(sources) {
         "Environment preflight must parse the retrieve-authored context.md baseUrl, reject unsafe/non-production violations, and curl-probe it before generate; unreachable or curl-unavailable ends preflight as blocked (frontend-base-url-unreachable|curl-unavailable) so generate/map do not run.",
         "U/D cases must prove current-user data ownership or create cleanable current-user data or authorized Mock; otherwise blocked (current-user-data-unavailable|data-ownership-unverifiable|safe-test-data-setup-unavailable) without cross-user mutation.",
         "Token settings are post-case stop thresholds, never a hard provider token cap. Unstarted cases after a threshold are blocked: token-budget-exhausted.",
-        "Pipeline acceptance for frontend-test is frontend-test-result-v1 plus the main report under testcase/frontend/reports/frontend-test-report.html; case pass rate and outcome=passed are quality signals; retrospect is opt-in (frontendTest.reports.retrospect).",
+        `Pipeline acceptance for frontend-test is frontend-test-result-v1 plus the main report under ${layout.reportsDir}/frontend-test-report.html; case pass rate and outcome=passed are quality signals; retrospect is opt-in (frontendTest.reports.retrospect).`,
         blockingReview
             ? "frontendTest.reviewMode=blocking: a frontend case review must emit VERDICT: pass before checklist/manifest materialization; request-revision blocks browser execution."
             : "frontendTest.reviewMode is off|advisory by default: mechanical checklist-shell gates materialize/execute; LLM review is not a hard browser gate.",
@@ -650,12 +651,32 @@ export function buildFrontendTestHybridDag(sources) {
         tasks,
     };
     applyFrontendTestLayoutToDagSpec(spec, layout);
+    applyFrontendTestHarnessRootGuard(spec, layout);
     applyDefaultReadOnlyRetryPolicy(spec);
     stampGeneratedArtifactBindings(spec);
     parseDagSpec(spec);
     assertValidDagSpec(spec);
     return spec;
 }
+/**
+ * D4: when the frozen frontend-test root lives under `.harness/` the whole tree
+ * is Git-ignored, so Git-status write guards would see empty diffs for real
+ * writes. Flip every file-writing node (and the dynamic browser child) from
+ * tools-only to the bounded filesystem snapshot channel instead.
+ */
+export function applyFrontendTestHarnessRootGuard(spec, layout) {
+    if (!isHarnessNamespacePath(layout.testRoot))
+        return;
+    for (const task of spec.tasks) {
+        if (task.writeGuardPolicy === "tools-only") {
+            task.writeGuardPolicy = "filesystem-only";
+        }
+        const child = task.dynamicExpansion?.childTask;
+        if (child?.writeGuardPolicy === "tools-only") {
+            child.writeGuardPolicy = "filesystem-only";
+        }
+    }
+}
 export function applyFrontendTestLayoutToDagSpec(spec, layout) {
     if (layout.isDefault) {
         spec.frontendTestLayout = layout;

package/dist/workflows/dag/test-artifact-namespace.js

@@ -0,0 +1,73 @@
+/**
+ * Shared test-artifact namespace policy (D3).
+ *
+ * Both dedicated test workflows (backend-test, frontend-test) default to a
+ * canonical root under `.harness/testcase/`. That root is the ONLY `.harness`
+ * subtree those workflows may write; every other `.harness/**` path stays
+ * fail-closed for controller-owned facts.
+ *
+ * This module is the single source of truth shared by:
+ * - frontend/backend layout resolvers (canonical default roots)
+ * - task source-prepare path assessment (precise ancestor-gate exception)
+ * - DAG validators / writeSet gates / writer tool policies / shell guards
+ *
+ * It must stay dependency-free (pure functions + constants only) so both
+ * `src/task` and `src/workflows` can import it without cycles.
+ */
+/** Canonical backend-test artifact root (repo-relative, posix). */
+export const HARNESS_TESTCASE_ROOT = ".harness/testcase";
+/** Canonical frontend-test artifact root (repo-relative, posix). */
+export const HARNESS_TESTCASE_FRONTEND_ROOT = `${HARNESS_TESTCASE_ROOT}/frontend`;
+/** taskKinds whose dedicated DAGs own the test-artifact namespace. */
+export const TEST_ARTIFACT_TASK_KINDS = [
+    "backend-test",
+    "frontend-test",
+];
+export function isTestArtifactTaskKind(value) {
+    return (typeof value === "string" &&
+        TEST_ARTIFACT_TASK_KINDS.includes(value));
+}
+/**
+ * Canonical test-artifact root for a dedicated test taskKind.
+ * backend-test -> `.harness/testcase`; frontend-test -> `.harness/testcase/frontend`.
+ */
+export function canonicalTestArtifactRoot(kind) {
+    return kind === "frontend-test"
+        ? HARNESS_TESTCASE_FRONTEND_ROOT
+        : HARNESS_TESTCASE_ROOT;
+}
+/**
+ * True when a repo-relative posix path is covered by the test-artifact
+ * allowance for the given kind (exact canonical root or beneath it).
+ */
+export function isTestArtifactPath(kind, repoRelativePosix) {
+    const root = canonicalTestArtifactRoot(kind);
+    return (repoRelativePosix === root ||
+        repoRelativePosix.startsWith(`${root}/`));
+}
+/**
+ * True when a repo-relative posix path is covered by EITHER dedicated test
+ * namespace (both kinds). Used by guards that authorize the exception but do
+ * not know which workflow owns the path.
+ */
+export function isAnyTestArtifactPath(repoRelativePosix) {
+    return (isTestArtifactPath("backend-test", repoRelativePosix) ||
+        isTestArtifactPath("frontend-test", repoRelativePosix));
+}
+/**
+ * True when a repo-relative posix path sits anywhere beneath `.harness/`
+ * (controller/runtime namespace) — regardless of test-artifact allowance.
+ */
+export function isHarnessNamespacePath(repoRelativePosix) {
+    return (repoRelativePosix === ".harness" ||
+        repoRelativePosix.startsWith(".harness/"));
+}
+/**
+ * True only for the baseline harness guard patterns (`.harness`, `.harness/**`).
+ * Deeper `.harness/<sub>/...` forbidden patterns are treated as explicit
+ * operator choices and stay authoritative over the test-artifact exception.
+ */
+export function isHarnessBaselineForbiddenPattern(pattern) {
+    const normalized = pattern.trim().replace(/\\/g, "/");
+    return normalized === ".harness" || normalized === ".harness/**";
+}

package/dist/workflows/dag/validate.js

@@ -5,6 +5,7 @@ import { resolveRepairTaskForGate } from "./repair-artifact.js";
 import { topoSortToRanks } from "./topo.js";
 import { isSafeReadOnlyPiRetryCandidate, isWriterEmptyDiffRetryCandidate, isWriterTransportRetryCandidate, INCOMPLETE_WRITE_SET_RETRY_CATEGORY, OUTPUT_LIMIT_RETRY_CATEGORY, WRITER_CLEAN_TIMEOUT_RETRY_CATEGORY, WRITER_EMPTY_DIFF_RETRY_CATEGORY, CONTEXT_OVERFLOW_RETRY_CATEGORY, } from "./retry-policy.js";
 import { isBackendTestCompletenessRetryCandidate } from "./backend-test-writer-completeness.js";
+import { isAnyTestArtifactPath, isHarnessBaselineForbiddenPattern, } from "./test-artifact-namespace.js";
 import { defaultFrontendTestLayout, frontendTestLayoutFromSpec, } from "./frontend-test-layout.js";
 const GOVERNANCE_WARNING_TYPES = new Set([
     "read-only-missing-artifacts-forbidden",
@@ -58,9 +59,27 @@ function writeSetEntryWithinAllowed(entry, allowedPaths) {
     const probes = probePathsForWriteSetEntry(entry);
     return probes.some((probe) => allowedPaths.some((allowed) => pathMatchesPattern(probe, allowed)));
 }
-function writeSetEntryHitsForbidden(entry, forbiddenPaths) {
+function writeSetEntryHitsForbidden(entry, forbiddenPaths, spec) {
     const probes = probePathsForWriteSetEntry(entry);
-    return probes.some((probe) => forbiddenPaths.some((forbidden) => pathMatchesPattern(probe, forbidden)));
+    // D3 test-artifact exception: entries covered by a frozen dedicated-test
+    // layout root under `.harness/testcase/` are the only `.harness` writes a
+    // test DAG may declare. The exception is exact-namespace (canonical roots
+    // only), so `.harness/foo` or `.harness/dag-runs/x` stays forbidden.
+    const frozenTestRoots = [
+        spec?.backendTestLayout?.testRoot,
+        spec?.frontendTestLayout?.testRoot,
+    ].filter((root) => typeof root === "string" && isAnyTestArtifactPath(root));
+    return probes.some((probe) => forbiddenPaths.some((forbidden) => {
+        if (frozenTestRoots.some((root) => probe === root || probe.startsWith(`${root}/`)) &&
+            pathMatchesPattern(probe, forbidden)) {
+            // probe is inside a canonical frozen test root: the BASELINE
+            // harness guard patterns (`.harness` / `.harness/**`) are
+            // exempted; any explicit deeper or custom matching pattern
+            // still wins.
+            return !isHarnessBaselineForbiddenPattern(forbidden);
+        }
+        return pathMatchesPattern(probe, forbidden);
+    }));
 }
 function forbiddenPathsExcludeRootArtifacts(forbiddenPaths) {
     return ROOT_ARTIFACT_PATH_PROBES.some((probe) => forbiddenPaths.some((forbidden) => pathMatchesPattern(probe, forbidden)));
@@ -381,7 +400,7 @@ function validateTaskWritePolicy(task, spec, issues) {
             continue;
         }
         if (forbiddenPaths.length > 0 &&
-            writeSetEntryHitsForbidden(entry, forbiddenPaths)) {
+            writeSetEntryHitsForbidden(entry, forbiddenPaths, spec)) {
             issues.push({
                 type: "write-set-forbidden",
                 message: `task ${task.id} writeSet entry "${entry}" matches forbiddenPaths`,

package/docs/templates/README.md

@@ -42,15 +42,15 @@
 - `frontend-test-dag.json` — frontend-test DAG 模板；其 deterministic checklist 会拒绝 alternative executable instructions，仅允许 allowlisted `playwright-cli` command。Pi generator/reviser 的 runtime writeSet 仅授权下列目标项目生成路径（不是本仓库文档索引目标）：
 
   ~~~text
-  testcase/frontend/cases/FE-*.md
-  testcase/frontend/cases/index.md
-  testcase/frontend/cases/manifest.draft.json
+  .harness/testcase/frontend/cases/FE-*.md
+  .harness/testcase/frontend/cases/index.md
+  .harness/testcase/frontend/cases/manifest.draft.json
   ~~~
 
   Pi writer writeSet 不包含 final manifest（文件名 manifest.json）。依赖 checklist 的 exclusive shell materializer 是 final manifest 的唯一写入者，以 temp+rename 原子生成 final，后续 map 只消费该 materializer 输出。
 - `frontend-test-dag.retrieve-context.prompt.md`、`frontend-test-dag.generate-cases.prompt.md`、`frontend-test-dag.review-cases.prompt.md`、`frontend-test-dag.review-execution.prompt.md`、`frontend-test-dag.retrospect.prompt.md` — 上下文、用例、审查、执行审查和复盘提示。
 - `frontend-test-case-checklist.md` — 前端测试用例检查清单。
-- `frontend-test-standard-scenarios.v1.json` — frontend-test 标准场景规范（generate 覆盖 must 场景；运行时可物化到目标项目 testcase/frontend/rag 下的 standard-scenarios.v1.json）。
+- `frontend-test-standard-scenarios.v1.json` — frontend-test 标准场景规范（generate 覆盖 must 场景；运行时可物化到目标项目 .harness/testcase/frontend/rag 下的 standard-scenarios.v1.json）。
 
 ## 知识同步与配置
 

package/docs/templates/backend-test-dag.generate-pytest.prompt.md

@@ -6,9 +6,9 @@
 
 ## 当前合同
 
-- 输入：最终 `testcase/md/**`、run-owned 环境报告与 Markdown 校验报告，以及有界的 pytest config / `conftest.py`。
+- 输入：最终 `.harness/testcase/md/**`、run-owned 环境报告与 Markdown 校验报告，以及有界的 pytest config / `conftest.py`。
 - 禁止重新读取 `source/**`、新增测试场景、重新分配 AC、执行 pytest、修改生产代码/配置或输出业务 JSON。
-- 写入范围：`testcase/**/test_*.py`、`testcase/**/helpers/**`、`testcase/**/factories/**`。
+- 写入范围：`.harness/testcase/**/test_*.py`、`.harness/testcase/**/helpers/**`、`.harness/testcase/**/factories/**`。
 - pytest 可以使用模块级 `test_*` 函数，也可以使用 pytest 测试类中的 `test_*` 方法；不得为了追溯强迫目标项目放弃现有 class-based 风格。
 - 每个最终 Markdown Case ID 必须出现在至少一个真实测试函数/方法区域中，优先同时出现在函数/方法名和 docstring：
 
@@ -26,7 +26,7 @@ class TestOrderApi:
 - 禁止 `skip` / `xfail`、吞断言、宽异常静默通过、mock 替代真实目标、删除用例或弱化断言。
 - best-effort 清理只能捕获所选 HTTP client 实际抛出的窄 transport exception，例如 `requests.RequestException` 或 `urllib.error.URLError`；禁止 `except:`、`except Exception`、`except BaseException` 后 `pass`。
 - 同一 Case ID 可以由多个 pytest 函数覆盖；额外映射会进入 traceability 报告，但不能伪造未在 Markdown 中定义的业务场景。
-- 生成 pytest 文件名必须与 Markdown 模块一一对应：`testcase/md/<module>.md` → `testcase/test_<module>.py`（`<module>` 为文件名去 `.md` 后小写、非字母数字转 `_`）。即使 Markdown `自动化映射` 写了别的路径，也必须写模块 stem 路径，不得额外发明 `test_be_*` 前缀。
+- 生成 pytest 文件名必须与 Markdown 模块一一对应：`.harness/testcase/md/<module>.md` → `.harness/testcase/test_<module>.py`（`<module>` 为文件名去 `.md` 后小写、非字母数字转 `_`）。即使 Markdown `自动化映射` 写了别的路径，也必须写模块 stem 路径，不得额外发明 `test_be_*` 前缀。
 - 测试失败必须诚实保留，后续节点只执行一次 pytest，并从同一 pytest-html 报告生成 HTML/facts。
 
 ## 推荐输出