fix: resolve 25 pre-publish blockers

- postinstall.mjs: fix alias package detection - migrate-legacy-plugin-entry: dedupe + regression tests - task_system: default consistency across runtime paths - task() contract: consistent tool behavior - runtime model selection, tool cap, stale-task cancellation - recovery sanitization, context-limit gating - Ralph semantic DONE hardening, Atlas fallback persistence - native-skill description/content, skill path traversal guard - publish workflow: platform awaited via reusable workflow job - release: version edits reapplied before commit/tag - JSONC plugin migration: top-level plugin key safety - cold-cache: user fallback models skip disconnected providers - docs/version/release framing updates Verified: bun test (4599 pass), tsc --noEmit clean, bun run build clean
2026-03-28 15:24:18 +09:00
parent 44b039bef6
commit d2c576c510
62 changed files with 1264 additions and 292 deletions
@@ -76,7 +76,9 @@ describe("resolveModelForDelegateTask", () => {
 		})

 		describe("#when availableModels is empty (cache exists but empty)", () => {
-			test("#then falls through to category default model (existing behavior)", () => {
+			test("#then keeps the category default when its provider is connected", () => {
+				const readConnectedProvidersSpy = spyOn(connectedProvidersCache, "readConnectedProvidersCache").mockReturnValue(["anthropic"])
+
 				const result = resolveModelForDelegateTask({
 					categoryDefaultModel: "anthropic/claude-sonnet-4-6",
 					fallbackChain: [
@@ -87,6 +89,40 @@ describe("resolveModelForDelegateTask", () => {
 				})

 				expect(result).toEqual({ model: "anthropic/claude-sonnet-4-6" })
+				readConnectedProvidersSpy.mockRestore()
+			})
+
+			test("#then skips a disconnected category default and resolves via a connected fallback", () => {
+				const readConnectedProvidersSpy = spyOn(connectedProvidersCache, "readConnectedProvidersCache").mockReturnValue(["openai"])
+
+				const result = resolveModelForDelegateTask({
+					categoryDefaultModel: "anthropic/claude-sonnet-4-6",
+					fallbackChain: [
+						{ providers: ["openai"], model: "gpt-5.4", variant: "high" },
+					],
+					availableModels: new Set(),
+					systemDefaultModel: "anthropic/claude-sonnet-4-6",
+				})
+
+				expect(result).toEqual({
+					model: "openai/gpt-5.4",
+					variant: "high",
+					fallbackEntry: { providers: ["openai"], model: "gpt-5.4", variant: "high" },
+					matchedFallback: true,
+				})
+				readConnectedProvidersSpy.mockRestore()
+			})
+
+			test("#then skips disconnected user fallback models and keeps the first connected fallback", () => {
+				const readConnectedProvidersSpy = spyOn(connectedProvidersCache, "readConnectedProvidersCache").mockReturnValue(["openai"])
+
+				const result = resolveModelForDelegateTask({
+					userFallbackModels: ["anthropic/claude-sonnet-4-6", "openai/gpt-5.4"],
+					availableModels: new Set(),
+				})
+
+				expect(result).toEqual({ model: "openai/gpt-5.4", matchedFallback: true })
+				readConnectedProvidersSpy.mockRestore()
 			})
 		})

@@ -225,16 +261,23 @@ describe("resolveModelForDelegateTask", () => {
 		})

 		describe("#when availableModels is empty", () => {
-			test("#then falls through to existing resolution (cache partially ready)", () => {
+			test("#then uses connected providers to avoid disconnected category defaults", () => {
+				const readConnectedProvidersSpy = spyOn(connectedProvidersCache, "readConnectedProvidersCache").mockReturnValue(["openai"])
+
 				const result = resolveModelForDelegateTask({
 					categoryDefaultModel: "anthropic/claude-sonnet-4-6",
 					fallbackChain: [
-						{ providers: ["anthropic"], model: "claude-sonnet-4-6" },
+						{ providers: ["openai"], model: "gpt-5.4" },
 					],
 					availableModels: new Set(),
 				})

-				expect(result).toBeDefined()
+				expect(result).toEqual({
+					model: "openai/gpt-5.4",
+					fallbackEntry: { providers: ["openai"], model: "gpt-5.4" },
+					matchedFallback: true,
+				})
+				readConnectedProvidersSpy.mockRestore()
 			})
 		})
 	})
@@ -59,6 +59,8 @@ export function resolveModelForDelegateTask(input: {
    return { model: userModel }
  }

+  const connectedProviders = input.availableModels.size === 0 ? readConnectedProvidersCache() : null
+
  // Before provider cache is created (first run), skip model resolution entirely.
  // OpenCode will use its system default model when no model is specified in the prompt.
  if (input.availableModels.size === 0 && !hasProviderModelsCache() && !hasConnectedProvidersCache()) {
@@ -77,7 +79,15 @@ export function resolveModelForDelegateTask(input: {
    }

    if (input.availableModels.size === 0) {
-      return { model: categoryDefault }
+      const categoryProvider = categoryDefault.includes("/") ? categoryDefault.split("/")[0] : undefined
+      if (!connectedProviders || !categoryProvider || connectedProviders.includes(categoryProvider)) {
+        return { model: categoryDefault }
+      }
+
+      log("[resolveModelForDelegateTask] skipping disconnected category default on cold cache", {
+        categoryDefault,
+        connectedProviders,
+      })
    }

    const parts = categoryDefault.split("/")
@@ -95,9 +105,19 @@ export function resolveModelForDelegateTask(input: {
  const userFallbackModels = input.userFallbackModels
  if (userFallbackModels && userFallbackModels.length > 0) {
    if (input.availableModels.size === 0) {
-      const first = userFallbackModels[0] ? parseUserFallbackModel(userFallbackModels[0]) : undefined
-      if (first) {
-        return { model: first.baseModel, variant: first.variant, matchedFallback: true }
+      for (const fallbackModel of userFallbackModels) {
+        const parsedFallback = parseUserFallbackModel(fallbackModel)
+        if (!parsedFallback) continue
+
+        if (
+          connectedProviders &&
+          parsedFallback.providerHint &&
+          !parsedFallback.providerHint.some((provider) => connectedProviders.includes(provider))
+        ) {
+          continue
+        }
+
+        return { model: parsedFallback.baseModel, variant: parsedFallback.variant, matchedFallback: true }
      }
    } else {
      for (const fallbackModel of userFallbackModels) {
@@ -115,7 +135,6 @@ export function resolveModelForDelegateTask(input: {
  const fallbackChain = input.fallbackChain
  if (fallbackChain && fallbackChain.length > 0) {
    if (input.availableModels.size === 0) {
-      const connectedProviders = readConnectedProvidersCache()
      if (connectedProviders) {
        const connectedSet = new Set(connectedProviders)
        for (const entry of fallbackChain) {
@@ -76,13 +76,13 @@ export function createDelegateTask(options: DelegateTaskToolOptions): ToolDefini
  - category: For task delegation (uses Sisyphus-Junior with category-optimized model)
  - subagent_type: For direct agent invocation (explore, librarian, oracle, etc.)
  
-  **DO NOT provide both.** If category is provided, subagent_type is ignored.
+  **DO NOT provide both.** category and subagent_type are mutually exclusive.
  
  - load_skills: ALWAYS REQUIRED. Pass [] if no skills needed, or ["skill-1", "skill-2"] for category tasks.
  - category: Use predefined category → Spawns Sisyphus-Junior with category config
    Available categories:
  ${categoryList}
-  - subagent_type: Use specific agent directly (explore, librarian, oracle, metis, momus)
+  - subagent_type: Use a specific callable non-primary agent directly (for example: explore, librarian, oracle, metis, momus)
  - run_in_background: REQUIRED. true=async (returns task_id), false=sync (waits). Use background=true ONLY for parallel exploration with 5+ independent queries.
  - session_id: Existing Task session to continue (from previous task output). Continues agent with FULL CONTEXT PRESERVED - saves tokens, maintains continuity.
  - command: The command that triggered this task (optional, for slash command tracking).
@@ -102,7 +102,7 @@ export function createDelegateTask(options: DelegateTaskToolOptions): ToolDefini
      prompt: tool.schema.string().describe("Full detailed prompt for the agent"),
      run_in_background: tool.schema.boolean().describe("REQUIRED. true=async (returns task_id), false=sync (waits). Use false for task delegation, true ONLY for parallel exploration."),
      category: tool.schema.string().optional().describe(`REQUIRED if subagent_type not provided. Do NOT provide both category and subagent_type.`),
-      subagent_type: tool.schema.string().optional().describe("REQUIRED if category not provided. Do NOT provide both category and subagent_type. Valid values: explore, librarian, oracle, metis, momus"),
+      subagent_type: tool.schema.string().optional().describe("REQUIRED if category not provided. Do NOT provide both category and subagent_type. Must be a callable non-primary agent name returned by app.agents()."),
      session_id: tool.schema.string().optional().describe("Existing Task session to continue"),
      command: tool.schema.string().optional().describe("The command that triggered this task"),
    },
@@ -115,7 +115,7 @@ export function createDelegateTask(options: DelegateTaskToolOptions): ToolDefini
          `  - You provided: category="${args.category}", subagent_type="${args.subagent_type}"\n` +
          `  - Use category for task delegation (e.g., category="${categoryExamples.split(", ")[0]}")\n` +
          `  - Use subagent_type for direct agent invocation (e.g., subagent_type="explore")\n` +
-          `  - Valid subagent_type values: explore, librarian, oracle, metis, momus`
+          `  - subagent_type must be a callable non-primary agent name returned by app.agents()`
        )
      }
      if (args.category) {