@tea-agent/loop-agent 0.39.0-beta.5 → 0.39.0-beta.7

package/dist/workflows/dag/backend-test-markdown-workflow.js

@@ -200,14 +200,170 @@ export async function inspectBackendTestEnvironment(input) {
         serverEntries,
     };
 }
+/**
+ * Filenames declared by the canonical `## Module Index` section of the backend
+ * README. Only a unique exact `## Module Index` heading is authoritative; a
+ * missing/duplicate section yields an empty declaration set (callers then fall
+ * back to the whole directory so legacy workspaces without a manifest keep
+ * their prior behaviour). Reads `[x](./x.md)`, `` `testcase/md/<x>.md` `` and
+ * inline `x.md` tokens inside the section; README.md itself is never a module.
+ */
+export function extractDeclaredModuleIndexFilenames(readme) {
+    const lines = readme
+        .replaceAll("\r\n", "\n")
+        .replaceAll("\r", "\n")
+        .split("\n");
+    const headings = lines.flatMap((line, index) => line.trim() === "## Module Index" ? [index] : []);
+    if (headings.length !== 1)
+        return [];
+    const start = headings[0] + 1;
+    const relativeEnd = lines
+        .slice(start)
+        .findIndex((line) => /^##\s+\S/.test(line.trim()));
+    const end = relativeEnd < 0 ? lines.length : start + relativeEnd;
+    const section = lines.slice(start, end).join("\n");
+    return unique([...section.matchAll(/`?[A-Za-z0-9_.-]+\.md`?/g)]
+        .map((match) => match[0].replaceAll("`", ""))
+        .filter((name) => name.toLowerCase() !== "readme.md"));
+}
+/**
+ * Count exact `## Module Index` headings in the backend README.
+ */
+function moduleIndexHeadingCountIn(readme) {
+    return readme
+        .replaceAll("\r\n", "\n")
+        .replaceAll("\r", "\n")
+        .split("\n")
+        .filter((line) => line.trim() === "## Module Index").length;
+}
+/**
+ * Machine-readable fail-closed finding for an invalid backend-test Module Index.
+ * These are distinct states (missing/unreadable/duplicate/empty-unparsable) so a
+ * stripped Module Index can never silently fall back to a whole-directory scan.
+ */
+export function backendManifestStatusFinding(manifestStatus, moduleIndexHeadingCount = 0) {
+    switch (manifestStatus) {
+        case "missing":
+            return "README is missing; backend-test current round fails closed without a declarative Module Index";
+        case "unreadable":
+            return "README is unreadable; backend-test current round fails closed without a usable Module Index";
+        case "duplicate":
+            return moduleIndexHeadingCount === 0
+                ? "README Module Index heading is missing; backend-test current round fails closed without a declarative Module Index"
+                : `duplicate ## Module Index heading (${moduleIndexHeadingCount} occurrences); backend-test current round fails closed without a unique Module Index`;
+        case "empty-unparsable":
+            return "README Module Index is empty/unparsable; backend-test current round fails closed with no declared module";
+        case "valid":
+            return "";
+    }
+}
+/**
+ * Resolve the current-round Markdown module manifest. The README `## Module
+ * Index` is the authoritative source for which module files belong to this
+ * round. Fail closed on every non-`valid` state (missing/unreadable/duplicate/
+ * empty-unparsable): the backend-test current round must NEVER re-scan the whole
+ * directory, or historical `.md` files could leak back into Case/coverage/
+ * partition facts.
+ *
+ * - `manifestStatus==="valid"`   — unique heading with >= 1 declared `.md` file.
+ * - `missing`                     — README file does not exist.
+ * - `unreadable`                  — README exists but could not be read.
+ * - `duplicate`                   — README has 0 or duplicate `## Module Index`.
+ * - `empty-unparsable`            — unique heading but no `.md` filename parsed.
+ * - `declaredMdFiles`  — declared filenames that exist on disk (relative paths).
+ * - `declaredMissing`  — declared filenames with no file on disk (fail-closed
+ *   evidence, never silently dropped).
+ * - `unmanifested`     — existing `.md` files that the Module Index does not
+ *   declare (non-silent advisory; they must not enter Case/coverage/partition).
+ */
+export async function resolveBackendTestManifestModules(workspaceRoot, layout) {
+    const resolved = layout ?? resolveBackendTestLayout(undefined);
+    const readmePath = path.join(workspaceRoot, ...resolved.readmePath.split("/"));
+    const invalid = (manifestStatus, moduleIndexHeadingCount) => ({
+        hasManifest: false,
+        manifestStatus,
+        moduleIndexHeadingCount,
+        declaredStems: [],
+        declaredMdFiles: [],
+        declaredMissing: [],
+        unmanifested: [],
+    });
+    let readme = "";
+    if (await exists(readmePath)) {
+        try {
+            readme = await readFile(readmePath, "utf8");
+        }
+        catch {
+            // Unreadable README is fail-closed (not silently treated as no manifest).
+            return invalid("unreadable", 0);
+        }
+    }
+    else {
+        return invalid("missing", 0);
+    }
+    const moduleIndexHeadingCount = moduleIndexHeadingCountIn(readme);
+    if (moduleIndexHeadingCount !== 1) {
+        return invalid("duplicate", moduleIndexHeadingCount);
+    }
+    const declaredStems = extractDeclaredModuleIndexFilenames(readme);
+    if (declaredStems.length === 0) {
+        return invalid("empty-unparsable", moduleIndexHeadingCount);
+    }
+    const hasManifest = true;
+    const directory = path.join(workspaceRoot, ...resolved.markdownDir.split("/"));
+    let existingNames = [];
+    try {
+        existingNames = (await readdir(directory, { withFileTypes: true }))
+            .filter((entry) => entry.isFile() &&
+            entry.name.toLowerCase().endsWith(".md") &&
+            entry.name.toLowerCase() !== "readme.md")
+            .map((entry) => entry.name);
+    }
+    catch {
+        // missing markdown directory -> no existing modules
+    }
+    const declaredByLower = new Map();
+    for (const name of declaredStems) {
+        const base = path.posix.basename(name);
+        declaredByLower.set(base.toLowerCase(), base);
+    }
+    const declaredMdFiles = [];
+    const declaredMissing = [];
+    for (const base of declaredByLower.values()) {
+        const disk = existingNames.find((name) => name.toLowerCase() === base.toLowerCase());
+        if (disk)
+            declaredMdFiles.push(`${resolved.markdownDir}/${disk}`);
+        else
+            declaredMissing.push(`${resolved.markdownDir}/${base}`);
+    }
+    const unmanifested = existingNames
+        .filter((name) => !declaredByLower.has(name.toLowerCase()))
+        .map((name) => `${resolved.markdownDir}/${name}`);
+    return {
+        hasManifest,
+        manifestStatus: "valid",
+        moduleIndexHeadingCount,
+        declaredStems,
+        declaredMdFiles: declaredMdFiles.sort(),
+        declaredMissing: declaredMissing.sort(),
+        unmanifested: unmanifested.sort(),
+    };
+}
 export async function markdownFiles(root, layout) {
     const resolved = layout ?? resolveBackendTestLayout(undefined);
     const directory = path.join(root, ...resolved.markdownDir.split("/"));
     const entries = await readdir(directory, { withFileTypes: true });
-    return entries
+    const all = entries
         .filter((entry) => entry.isFile() && entry.name.toLowerCase().endsWith(".md"))
         .map((entry) => path.join(directory, entry.name))
         .sort();
+    const manifest = await resolveBackendTestManifestModules(root, resolved);
+    if (manifest.manifestStatus !== "valid")
+        return [];
+    const declaredAbs = new Set(manifest.declaredMdFiles.map((rel) => path.join(root, ...rel.split("/")).toLowerCase()));
+    return all
+        .filter((file) => declaredAbs.has(file.toLowerCase()))
+        .sort();
 }
 function canonicalBackendCaseId(value) {
     const upper = value.toUpperCase();
@@ -453,6 +609,17 @@ export async function validateBackendMarkdownCases(input) {
     const readme = path.join(input.workspaceRoot, ...layout.readmePath.split("/"));
     if (!(await exists(readme)))
         findings.push(`missing ${layout.readmePath}`);
+    const manifest = await resolveBackendTestManifestModules(input.workspaceRoot, layout);
+    // Fail closed on README Module Index entries that were declared but never
+    // materialized — never let a declared module silently vanish from the round.
+    if (manifest.declaredMissing.length > 0) {
+        findings.push(`declared Markdown module is missing (Module Index): ${manifest.declaredMissing.join(", ")}`);
+    }
+    // Fail closed on every non-valid Module Index state: without a valid unique
+    // Module Index the current round must not re-scan historical `.md` files.
+    if (manifest.manifestStatus !== "valid") {
+        findings.push(backendManifestStatusFinding(manifest.manifestStatus, manifest.moduleIndexHeadingCount));
+    }
     const moduleFiles = files.filter((file) => path.basename(file).toLowerCase() !== "readme.md");
     if (moduleFiles.length === 0)
         findings.push(`no ${layout.markdownDir} module case file found`);
@@ -562,6 +729,7 @@ export async function validateBackendMarkdownCases(input) {
         `- Cases: ${caseCount}`,
         `- Covered AC IDs: ${[...coveredAc].sort().join(", ") || "none"}`,
         `- Environment evidence present: ${input.environmentMarkdown.trim() ? "yes" : "no"}`,
+        `- Manifest scope: ${manifest.manifestStatus === "valid" ? "declared-modules" : `fails-closed:${manifest.manifestStatus}`} (declared-missing=${manifest.declaredMissing.length}, unmanifested-advisory=${manifest.unmanifested.length}, module-index-heading-count=${manifest.moduleIndexHeadingCount})`,
         `- Findings: ${findings.length}`,
         "- Required sections: Preconditions/前置条件, Steps/操作步骤, Expected Results/预期结果",
         "- Source-reference validity check: not applied",
@@ -583,6 +751,13 @@ export async function validateBackendMarkdownCases(input) {
             ? safetyFindings.map((finding) => `- ${finding}`)
             : ["- No secret-shaped values detected."]),
         "",
+        "## Unmanifested Markdown Modules",
+        "",
+        "- Advisory: existing .md files not declared by the README Module Index are excluded from this round's Case/coverage/partition scope.",
+        ...(manifest.unmanifested.length > 0
+            ? manifest.unmanifested.map((item) => `- ${item}`)
+            : ["- None"]),
+        "",
         "## Findings",
         "",
         ...(findings.length > 0 ? findings.map((finding) => `- ${finding}`) : ["- None"]),

package/dist/workflows/dag/backend-test-pytest-collection.js

@@ -131,6 +131,23 @@ function normalizeHazardOccurrenceIdentity(line) {
     // Unclosed quote or no out-of-string `#`: do not invent a comment boundary.
     return trimmed;
 }
+/**
+ * Bare `__NNN_*` helper module names imported by a Python source snippet.
+ * Only the head token of an `import`/`from` statement is considered a module
+ * name; dotted submodules (`__551_manager_helpers.x`) collapse to the head.
+ */
+function importedBareHelperModules(source) {
+    const names = [];
+    const seen = new Set();
+    for (const match of normalizeSourceNewlines(source).matchAll(/\b(?:import|from)\s+([A-Za-z0-9_.]+)/g)) {
+        const head = match[1].split(".")[0];
+        if (/^__\d+[A-Za-z0-9_]*$/.test(head) && !seen.has(head)) {
+            seen.add(head);
+            names.push(head);
+        }
+    }
+    return names;
+}
 function listPatternOccurrences(source, patternSource, flags = "") {
     const normalized = normalizeSourceNewlines(source);
     const pattern = new RegExp(patternSource, flags.includes("g") ? flags : `${flags}g`);
@@ -221,6 +238,8 @@ export async function buildBackendPytestAssetInventory(workspaceRoot, mappedScri
     const inputHashes = {};
     const hazardOccurrenceCounts = {};
     const hazardOccurrences = {};
+    const scriptBareImports = {};
+    const mappedScriptSet = new Set(normalizedScripts);
     for (const file of assetFiles) {
         const bytes = await readFile(path.resolve(workspaceRoot, file));
         const source = bytes.toString("utf8");
@@ -228,8 +247,10 @@ export async function buildBackendPytestAssetInventory(workspaceRoot, mappedScri
         const lists = scanRepairHazardOccurrenceLists(source);
         hazardOccurrences[file] = lists;
         hazardOccurrenceCounts[file] = Object.fromEntries(Object.entries(lists).map(([label, identities]) => [label, identities.length]));
+        if (mappedScriptSet.has(file))
+            scriptBareImports[file] = importedBareHelperModules(source);
     }
-    return { mappedScripts: normalizedScripts, assetFiles, inputHashes, hazardOccurrenceCounts, hazardOccurrences };
+    return { mappedScripts: normalizedScripts, assetFiles, inputHashes, hazardOccurrenceCounts, hazardOccurrences, scriptBareImports };
 }
 function bounded(value) {
     const normalized = value.replaceAll(/(Bearer|Basic)\s+[A-Za-z0-9._~+\/-]+/gi, "$1 [REDACTED]");
@@ -308,8 +329,68 @@ function classifyFixtureResolutionFailure(output, layout) {
         repairPaths: paths,
     };
 }
-function classifyCollectionFailure(output) {
+/**
+ * Generated-local bare helper module signatures found in the bound pytest
+ * asset inventory. A file such as `testcase/__551_manager_helpers.py` in the
+ * test root/script dir is generated-local: when collection reports
+ * `No module named '__551_manager_helpers'`, the defect can be precisely
+ * attributed to the generated mapped script instead of a missing third-party
+ * package.
+ */
+function generatedLocalBareHelperNames(inventory) {
+    if (!inventory)
+        return [];
+    const names = [];
+    for (const file of inventory.assetFiles) {
+        const stem = path.posix.basename(file).replace(/\.py$/, "");
+        if (/^__\d+[A-Za-z0-9_]*$/.test(stem))
+            names.push(stem);
+    }
+    return [...new Set(names)];
+}
+function classifyCollectionFailure(output, evidence) {
     const normalized = output.replaceAll("\\", "/");
+    const helperNames = generatedLocalBareHelperNames(evidence?.inventory);
+    const helperByLower = new Map(helperNames.map((name) => [name.toLowerCase(), name]));
+    // A bare `__NNN_...` module import that collection cannot find is only a
+    // candidate generated-local helper signature. REPAIRABLE requires two
+    // deterministic pieces of evidence: (1) a helper asset of the same name in
+    // the bound inventory, and (2) at least one mapped generated script whose
+    // source truly imports that bare module. Without both — or when attribution
+    // is impossible — fail closed as BLOCKED so an un-attributable dependency
+    // error is never widened into a repair. Genuine third-party modules (no
+    // `__NNN_` signature) stay BLOCKED below.
+    const missingModule = /No module named ['\"]([^'\"]+)['\"]/i.exec(normalized)?.[1];
+    if (missingModule && /^__\d+[A-Za-z0-9_]*$/.test(missingModule)) {
+        const helperAsset = helperByLower.get(missingModule.toLowerCase());
+        const inventory = evidence?.inventory;
+        const scriptBareImports = inventory?.scriptBareImports ?? {};
+        const missingLower = missingModule.toLowerCase();
+        const importedByMappedScript = inventory
+            ? inventory.mappedScripts.some((script) => (scriptBareImports[script] ?? []).some((name) => name.toLowerCase() === missingLower))
+            : false;
+        if (helperAsset && importedByMappedScript && inventory) {
+            const repairPaths = [...new Set([
+                    ...(helperAsset
+                        ? inventory.assetFiles.filter((file) => path.posix.basename(file).replace(/\.py$/, "") === helperAsset)
+                        : []),
+                    ...inventory.mappedScripts,
+                ])].sort();
+            return {
+                status: "REPAIRABLE",
+                kind: "missing-generated-local-helper",
+                detail: `generated testcase asset imports missing bare generated-local helper module ${missingModule}; inline the helper or remove the generated shared dependency`,
+                repairPaths,
+            };
+        }
+        return {
+            status: "BLOCKED",
+            kind: "missing-generated-local-helper-without-evidence",
+            detail: helperAsset
+                ? `missing bare generated-local helper module ${missingModule} has a name match but no mapped generated script imports it (no precise generated-local attribution)`
+                : `missing bare __NNN_* module ${missingModule} has no generated-local helper asset in the bound inventory (cannot be precisely attributed)`,
+        };
+    }
     const blockedPatterns = [
         [/No module named ['\"](?!testcase(?:\.|['\"]))/i, "missing-third-party-module"],
         [/pytest(?:_|-)plugin|pluggy|unknown hook|unrecognized arguments/i, "pytest-plugin-or-config"],
@@ -401,7 +482,7 @@ export function assessBackendPytestCollection(input) {
             fixtureStderrExcerpt: bounded(fixtureResolution.stderr),
         });
     }
-    const classification = classifyCollectionFailure(`${input.stdout}\n${input.stderr}`);
+    const classification = classifyCollectionFailure(`${input.stdout}\n${input.stderr}`, { layout: input.layout, inventory: input.inventory });
     return backendPytestCollectionFactsSchema.parse({
         schemaId: "backend-test-pytest-collection-v3",
         phase: input.phase,
@@ -412,7 +493,9 @@ export function assessBackendPytestCollection(input) {
         fixtureResolutionAttempted: false,
         fixtureResolutionStatus: "NOT_RUN",
         fixtureResolutionExitCode: null,
-        repairPaths: testcasePythonPaths(`${input.stdout}\n${input.stderr}`),
+        repairPaths: classification.repairPaths && classification.repairPaths.length > 0
+            ? classification.repairPaths
+            : testcasePythonPaths(`${input.stdout}\n${input.stderr}`),
         mappedScripts: input.inventory.mappedScripts,
         existingMappedScripts: input.inventory.mappedScripts,
         missingMappedScripts: [],

package/dist/workflows/dag/backend-test-writer-completeness.js

@@ -385,12 +385,17 @@ export const FORBIDDEN_GENERATED_IMPORT_NAMESPACES = [
     "testcase.helpers",
     "testcase.factories",
 ];
-const FORBIDDEN_GENERATED_IMPORT_PATTERN = /^\s*(?:from\s+(testcase_helpers|testcase\.helpers|testcase\.factories)\b|import\s+(testcase_helpers|testcase\.helpers|testcase\.factories)\b)/gm;
+const FORBIDDEN_GENERATED_IMPORT_PATTERN = /^\s*(?:from\s+(testcase_helpers|testcase\.helpers|testcase\.factories|__\d+[A-Za-z0-9_]*)\b|import\s+(testcase_helpers|testcase\.helpers|testcase\.factories|__\d+[A-Za-z0-9_]*)\b)/gm;
 /**
- * Scan Python source for imports from forbidden generated shared namespaces.
- * Anchored at line start with a word boundary so `testcase.helpers_other` or a
- * substring inside a docstring sentence is not falsely flagged. Returns the
- * list of offending namespace module names (deduplicated, order-preserving).
+ * Scan Python source for imports from forbidden generated shared namespaces
+ * or bare generated-local helper modules. Anchored at line start with a word
+ * boundary so `testcase.helpers_other` or a substring inside a docstring is not
+ * falsely flagged. In addition to the `testcase_helpers`/`testcase.helpers`/
+ * `testcase.factories` namespaces, bare `__<digits>...` helper modules (for
+ * example `from __551_manager_helpers import ...` / `import __551_manager_helpers`)
+ * are generated-local shared dependencies: a self-contained pytest child must
+ * inline or remove them rather than depend on a generated shared module.
+ * Returns the list of offending module names (deduplicated, order-preserving).
  */
 export function findForbiddenGeneratedImports(source) {
     const hits = [];

package/dist/workflows/dag/dynamic-runtime/loop-until.js

@@ -76,7 +76,7 @@ export async function executeDynamicLoopUntil(input) {
             bodyIdMap,
         }));
         for (const child of children) {
-            assertValidMaterializedDagTask(child);
+            assertValidMaterializedDagTask(child, input.spec);
             if (!input.tasksById.has(child.id)) {
                 input.tasksById.set(child.id, child);
                 input.spec.tasks.push(child);

package/dist/workflows/dag/dynamic-runtime/map.js

@@ -170,7 +170,7 @@ export async function executeDynamicMapExpansion(input) {
         nodeId,
     }));
     for (const child of children) {
-        assertValidMaterializedDagTask(child);
+        assertValidMaterializedDagTask(child, input.spec);
     }
     const writeSetConflicts = collectDynamicChildWriteSetConflicts(children);
     if (writeSetConflicts.length > 0) {

package/dist/workflows/dag/failure-category.js

@@ -1,130 +1,9 @@
 import { z } from "zod";
-export const dagNormalizedFailureCategorySchema = z.enum([
-    "success",
-    "validation",
-    "executor",
-    "write-guard",
-    "timeout",
-    "auth",
-    "human-required",
-    "human-rejected",
-    "shell-command",
-    "static-error",
-    "decision-envelope",
-    "skipped",
-    "unknown",
-]);
+import { DAG_NORMALIZED_FAILURE_CATEGORY_VALUES } from "../../shared/dag-failure-category.js";
+// The pure classification core lives in src/shared/dag-failure-category.ts so
+// the Operator Console projection can normalize raw CLI categories without
+// importing the workflows kernel (console/workflows boundary test). This
+// module keeps the schema surface for the DAG kernel's existing consumers.
+export { isDagFailureCategory, normalizeDagFailureCategory, } from "../../shared/dag-failure-category.js";
+export const dagNormalizedFailureCategorySchema = z.enum(DAG_NORMALIZED_FAILURE_CATEGORY_VALUES);
 export const DAG_NORMALIZED_FAILURE_CATEGORIES = dagNormalizedFailureCategorySchema.options;
-const RAW_TO_NORMALIZED = {
-    success: "success",
-    "write-guard": "write-guard",
-    timeout: "timeout",
-    // Clean implementer stall with zero writes — still a timeout class for report,
-    // but retryable under WRITER_TRANSPORT_RETRY_POLICY.
-    "writer-clean-timeout": "timeout",
-    "missing-api-key": "auth",
-    auth: "auth",
-    "human-rejected": "human-rejected",
-    "human-required": "human-required",
-    "decision-envelope-invalid": "decision-envelope",
-    "nonzero-exit": "shell-command",
-    "spawn-error": "shell-command",
-    "test-timeout": "timeout",
-    "static-error": "static-error",
-    "invalid-output": "validation",
-    "structured-output-truncated": "validation",
-    "output-too-large": "validation",
-    "structured-repair-exhausted": "validation",
-    "governance-blocked": "validation",
-    "path-violation": "validation",
-    "empty-output": "validation",
-    "verify-failure": "validation",
-    "test-failure": "validation",
-    "verification-plan-stale": "validation",
-    network: "executor",
-    quota: "executor",
-    "rate-limit": "executor",
-    unavailable: "executor",
-    // Deterministic capability/policy refusal: missing verified launcher JS
-    // entry, capability allowlist rejection, or missing bounded write tool
-    // profile. These are environment/config defects the operator repairs
-    // before rerunning; classify as executor so recovery recommends
-    // rerun-after-fix WITHOUT forcing a human-decision gate on
-    // standaloneTaskRerun (run-failed same-task rerun).
-    "tool-policy": "executor",
-    unknown: "unknown",
-};
-function isFinishedStatus(status) {
-    return status === "FINISHED" || status === "finished";
-}
-function isInProgressStatus(status) {
-    return (status === "PENDING" ||
-        status === "RUNNING" ||
-        status === "pending" ||
-        status === "running");
-}
-function isFailureStatus(status) {
-    return (status === "ERROR" ||
-        status === "failed" ||
-        status === "partial_failed");
-}
-function mapRawCategory(raw) {
-    const direct = RAW_TO_NORMALIZED[raw];
-    if (direct)
-        return direct;
-    const lower = raw.toLowerCase();
-    if (lower.includes("timeout"))
-        return "timeout";
-    if (lower.includes("validation") ||
-        lower.includes("invalid") ||
-        lower.includes("verification-plan-stale") ||
-        lower.includes("plan-stale"))
-        return "validation";
-    if (lower.includes("auth") || lower.includes("api-key"))
-        return "auth";
-    if (lower.includes("write-guard") || lower.includes("write_guard"))
-        return "write-guard";
-    if (lower.includes("decision-envelope") || lower.includes("decision_envelope"))
-        return "decision-envelope";
-    if (lower.includes("human-rejected"))
-        return "human-rejected";
-    if (lower.includes("human-required"))
-        return "human-required";
-    if (lower.includes("static-error"))
-        return "static-error";
-    if (lower.includes("nonzero-exit") ||
-        lower.includes("spawn-error") ||
-        lower.includes("shell"))
-        return "shell-command";
-    return undefined;
-}
-export function isDagFailureCategory(value) {
-    return DAG_NORMALIZED_FAILURE_CATEGORIES.includes(value);
-}
-export function normalizeDagFailureCategory(raw, status) {
-    if (status === "SKIPPED") {
-        return "skipped";
-    }
-    if (raw !== undefined) {
-        const mapped = mapRawCategory(raw);
-        if (mapped) {
-            if (mapped === "success") {
-                return isFinishedStatus(status) ? "success" : "unknown";
-            }
-            return mapped;
-        }
-    }
-    if (raw === undefined && isFinishedStatus(status)) {
-        return "success";
-    }
-    if (isInProgressStatus(status)) {
-        return "unknown";
-    }
-    if (isFailureStatus(status)) {
-        return "unknown";
-    }
-    if (raw === "success" && isFinishedStatus(status)) {
-        return "success";
-    }
-    return "unknown";
-}

package/dist/workflows/dag/frontend-design-policy.js

@@ -2,6 +2,10 @@ import { deterministicSha256, } from "./frontend-implementation-contract.js";
 import { pathMatchesPattern } from "../../shared/git-progress.js";
 const SHA256_HEX = /^[a-f0-9]{64}$/;
 const NPM_PACKAGE_NAME = /^(?:@[a-zA-Z0-9_.-]+\/)?[a-zA-Z0-9][a-zA-Z0-9_.-]*$/;
+/** Max concrete implementation files a single writeSet may carry before the
+ * plan is asked to split. Kept at 15: beyond this, a single implement node's
+ * read/edit loop inflates provider context toward overflow. */
+const MAX_WRITE_SET_FILES = 15;
 function normalizeRelativePath(value) {
     return value.replace(/\\/g, "/").replace(/^\.\//, "");
 }
@@ -26,6 +30,11 @@ export function checkUiDesignCoverage(contract) {
     const findings = [];
     const purposes = new Set((contract.uiComponentChoices ?? []).map((choice) => choice.purpose));
     const hasStyling = Boolean(contract.stylingStrategy?.trim());
+    // Reuse of an existing component/pattern covers behavioural interactions
+    // without forcing one uiComponentChoices row per interaction: the model
+    // should not enumerate every click/hover on a reused Button. Only
+    // interactions with no reuse and no explicit purpose are flagged.
+    const hasReuseChoice = (contract.uiComponentChoices ?? []).some((choice) => choice.decision === "reuse-existing");
     for (const state of contract.uiStates) {
         if (!state.applicable)
             continue;
@@ -39,13 +48,13 @@ export function checkUiDesignCoverage(contract) {
         }
     }
     for (const interaction of contract.interactions) {
-        if (!purposes.has(interaction.name)) {
-            findings.push({
-                code: "ui-design-coverage-missing",
-                message: `interaction "${interaction.name}" has no matching uiComponentChoices.purpose`,
-                path: interaction.name,
-            });
-        }
+        if (purposes.has(interaction.name) || hasReuseChoice)
+            continue;
+        findings.push({
+            code: "ui-design-coverage-missing",
+            message: `interaction "${interaction.name}" has no matching uiComponentChoices.purpose and no reuse-existing choice covering it`,
+            path: interaction.name,
+        });
     }
     return findings;
 }
@@ -204,7 +213,22 @@ export function checkComponentChoice(contract, componentSpecCandidatePaths) {
  * must fall inside the declared writeSet patterns. */
 export function checkTargetPaths(contract, writeSetPatterns) {
     const findings = [];
+    // No frozen writeSet patterns (legacy / unconfigured run) → nothing to
+    // check; an empty pattern set must not report every file as out of scope.
+    if (writeSetPatterns.length === 0)
+        return findings;
     const matches = (file) => writeSetPatterns.some((pattern) => pathMatchesPattern(file, pattern));
+    // A writeSet this large forces a single implement node to read/edit many
+    // files, inflating the provider context until it overflows (400
+    // request-too-large). Flag it so the plan is asked to split (the contract
+    // node already has record_split_proposal) instead of letting the writer
+    // blow up mid-run.
+    if (contract.targets.files.length > MAX_WRITE_SET_FILES) {
+        findings.push({
+            code: "write-set-too-large",
+            message: `implementation writeSet has ${contract.targets.files.length} files (limit ${MAX_WRITE_SET_FILES}); split the task (record_split_proposal) so no single implement node exceeds the bound`,
+        });
+    }
     for (const file of contract.targets.files) {
         if (!matches(file)) {
             findings.push({