Merge pull request #3226 from code-yeongyu/fix/plan-progress-and-trust

fix(plan): skill MCP trust + non-ASCII names + simple-mode progress + ralph-loop cap

src/features/boulder-state/storage.test.ts

@@ -650,6 +650,65 @@ describe("boulder-state", () => {
       expect(progress.completed).toBe(1)
       expect(progress.isComplete).toBe(false)
     })
+
+    test("should count only top-level checkboxes for simple plans with nested tasks", () => {
+      // given
+      const planPath = join(TEST_DIR, "simple-nested-plan.md")
+      writeFileSync(planPath, `# Plan
+
+- [ ] Top-level task 1
+  - [x] Nested task ignored
+- [x] Top-level task 2
+    * [ ] Another nested task ignored
+`)
+
+      // when
+      const progress = getPlanProgress(planPath)
+
+      // then
+      expect(progress.total).toBe(2)
+      expect(progress.completed).toBe(1)
+      expect(progress.isComplete).toBe(false)
+    })
+
+    test("should treat final-wave-only plans as structured mode", () => {
+      // given
+      const planPath = join(TEST_DIR, "final-wave-only-plan.md")
+      writeFileSync(planPath, `# Plan
+
+## Final Verification Wave
+- [ ] F1. Top-level final review
+  - [x] Nested verification detail ignored
+`)
+
+      // when
+      const progress = getPlanProgress(planPath)
+
+      // then
+      expect(progress.total).toBe(1)
+      expect(progress.completed).toBe(0)
+      expect(progress.isComplete).toBe(false)
+    })
+
+    test("should ignore mixed indentation levels in simple plans", () => {
+      // given
+      const planPath = join(TEST_DIR, "simple-mixed-indentation-plan.md")
+      writeFileSync(planPath, `# Plan
+
+* [x] Top-level star task
+ - [ ] Indented task ignored
+	- [x] Tab-indented task ignored
+- [ ] Top-level dash task
+`)
+
+      // when
+      const progress = getPlanProgress(planPath)
+
+      // then
+      expect(progress.total).toBe(2)
+      expect(progress.completed).toBe(1)
+      expect(progress.isComplete).toBe(false)
+    })
   })
 
   describe("getPlanName", () => {

src/features/boulder-state/storage.ts

@@ -226,7 +226,9 @@ export function getPlanProgress(planPath: string): PlanProgress {
     const lines = content.split(/\r?\n/)
 
     // Check if the plan has structured sections (## TODOs / ## Final Verification Wave)
-    const hasStructuredSections = lines.some((line) => TODO_HEADING_PATTERN.test(line))
+    const hasStructuredSections = lines.some(
+      (line) => TODO_HEADING_PATTERN.test(line) || FINAL_VERIFICATION_HEADING_PATTERN.test(line),
+    )
 
     if (hasStructuredSections) {
       // Structured plan: only count top-level checkboxes with numbered labels
@@ -291,8 +293,8 @@ function getStructuredPlanProgress(lines: string[]): PlanProgress {
 }
 
 function getSimplePlanProgress(content: string): PlanProgress {
-  const uncheckedMatches = content.match(/^\s*[-*]\s*\[\s*\]/gm) || []
+  const uncheckedMatches = content.match(/^[-*]\s*\[\s*\]/gm) || []
-  const checkedMatches = content.match(/^\s*[-*]\s*\[[xX]\]/gm) || []
+  const checkedMatches = content.match(/^[-*]\s*\[[xX]\]/gm) || []
 
   const total = uncheckedMatches.length + checkedMatches.length
   const completed = checkedMatches.length

src/features/builtin-commands/templates/ralph-loop.test.ts

@@ -0,0 +1,15 @@
+import { describe, expect, test } from "bun:test"
+import { ULW_LOOP_TEMPLATE } from "./ralph-loop"
+
+describe("ULW_LOOP_TEMPLATE", () => {
+  test("returns the documented iteration caps for ultrawork and normal modes", () => {
+    // given
+    const expectedIterationCaps = "The iteration limit is 500 for ultrawork mode, 100 for normal mode"
+
+    // when
+    const template = ULW_LOOP_TEMPLATE
+
+    // then
+    expect(template).toContain(expectedIterationCaps)
+  })
+})

src/features/builtin-commands/templates/ralph-loop.ts

@@ -36,7 +36,7 @@ export const ULW_LOOP_TEMPLATE = `You are starting an ULTRAWORK Loop - a self-re
 2. When you believe the work is complete, output: \`<promise>{{COMPLETION_PROMISE}}</promise>\`
 3. That does NOT finish the loop yet. The system will require Oracle verification
 4. The loop only ends after the system confirms Oracle verified the result
-5. There is no iteration limit
+5. The iteration limit is 500 for ultrawork mode, 100 for normal mode
 
 ## Rules
 

src/features/skill-mcp-manager/connection-env-vars.test.ts

@@ -1,4 +1,4 @@
-import { afterAll, afterEach, beforeEach, describe, expect, it, mock } from "bun:test"
+import { afterAll, afterEach, beforeEach, describe, expect, it, mock, test } from "bun:test"
 import type { ClaudeCodeMcpServer } from "../claude-code-mcp-loader/types"
 import type { SkillMcpClientInfo, SkillMcpManagerState } from "./types"
 
@@ -89,12 +89,15 @@ function createState(): SkillMcpManagerState {
   return state
 }
 
-function createClientInfo(serverName: string): SkillMcpClientInfo {
+function createClientInfo(
+  serverName: string,
+  scope?: SkillMcpClientInfo["scope"],
+): SkillMcpClientInfo {
   return {
     serverName,
     skillName: "env-skill",
     sessionID: "session-env",
-    scope: "builtin",
+    ...(scope !== undefined ? { scope } : {}),
   }
 }
 
@@ -126,6 +129,68 @@ afterEach(async () => {
 })
 
 describe("getOrCreateClient env var expansion", () => {
+  describe("#given a scope-sensitive stdio skill MCP config", () => {
+    test.each([
+      ["opencode-project", "Authorization:Bearer "],
+      ["local", "Authorization:Bearer "],
+      ["user", "Authorization:Bearer xoxp-scope-token"],
+      ["builtin", "Authorization:Bearer xoxp-scope-token"],
+    ] satisfies Array<[NonNullable<SkillMcpClientInfo["scope"]>, string]>) (
+      "#when creating the client for %s scope #then args expand to %s",
+      async (scope, expectedAuthorizationHeader) => {
+        // given
+        process.env.SLACK_USER_TOKEN = "xoxp-scope-token"
+        const state = createState()
+        const info = createClientInfo(`scope-${scope}`, scope)
+        const clientKey = createClientKey(info)
+        const config: ClaudeCodeMcpServer = {
+          command: "npx",
+          args: [
+            "-y",
+            "mcp-remote",
+            "https://mcp.slack.com/mcp",
+            "--header",
+            "Authorization:Bearer ${SLACK_USER_TOKEN}",
+          ],
+        }
+
+        // when
+        await getOrCreateClient({ state, clientKey, info, config })
+
+        // then
+        expect(createdStdioTransports).toHaveLength(1)
+        expect(createdStdioTransports[0]?.options.args?.[4]).toBe(expectedAuthorizationHeader)
+      },
+    )
+
+    it("#when creating the client without scope #then env vars remain trusted for backward compatibility", async () => {
+      // given
+      process.env.SLACK_USER_TOKEN = "xoxp-undefined-scope-token"
+      const state = createState()
+      const info = createClientInfo("scope-undefined")
+      const clientKey = createClientKey(info)
+      const config: ClaudeCodeMcpServer = {
+        command: "npx",
+        args: [
+          "-y",
+          "mcp-remote",
+          "https://mcp.slack.com/mcp",
+          "--header",
+          "Authorization:Bearer ${SLACK_USER_TOKEN}",
+        ],
+      }
+
+      // when
+      await getOrCreateClient({ state, clientKey, info, config })
+
+      // then
+      expect(createdStdioTransports).toHaveLength(1)
+      expect(createdStdioTransports[0]?.options.args?.[4]).toBe(
+        "Authorization:Bearer xoxp-undefined-scope-token",
+      )
+    })
+  })
+
   describe("#given a stdio skill MCP config with sensitive env vars in args", () => {
     it("#when creating the client #then sensitive env vars in args are expanded", async () => {
       // given

src/features/skill-mcp-manager/connection.ts

@@ -14,6 +14,8 @@ function removeClientIfCurrent(state: SkillMcpManagerState, clientKey: string, c
   }
 }
 
+const PROJECT_SCOPES = new Set(["project", "opencode-project", "local"])
+
 export async function getOrCreateClient(params: {
   state: SkillMcpManagerState
   clientKey: string
@@ -38,7 +40,7 @@ export async function getOrCreateClient(params: {
     return pending
   }
 
-  const isTrusted = info.scope !== "project"
+  const isTrusted = !PROJECT_SCOPES.has(info.scope ?? "")
   const expandedConfig = expandEnvVarsInObject(config, { trusted: isTrusted })
   let currentConnectionPromise!: Promise<Client>
   state.inFlightConnections.set(info.sessionID, (state.inFlightConnections.get(info.sessionID) ?? 0) + 1)

src/features/skill-mcp-manager/types.ts

@@ -11,7 +11,7 @@ export interface SkillMcpClientInfo {
   serverName: string
   skillName: string
   sessionID: string
-  scope?: SkillScope
+  scope?: SkillScope | "local"
 }
 
 export interface SkillMcpServerContext {

src/hooks/start-work/context-info-builder.ts

@@ -23,7 +23,7 @@ function normalizePlanLookupValue(value: string): string {
     .replace(/^["'`]+|["'`]+$/g, "")
     .toLowerCase()
     .replace(/[\s_]+/g, "-")
-    .replace(/[^a-z0-9-]+/g, "-")
+    .replace(/[^\p{L}\p{N}-]+/gu, "-")
     .replace(/-+/g, "-")
     .replace(/^-+|-+$/g, "")
 }

src/hooks/start-work/index.test.ts

@@ -443,6 +443,122 @@ You are starting a Sisyphus work session.
       expect(output.parts[0].text).toContain("my-feature-plan")
       expect(output.parts[0].text).toContain("Auto-Selected Plan")
     })
+
+    test("should match Korean plan names after Unicode-aware normalization", async () => {
+      // given
+      const plansDir = join(testDir, ".sisyphus", "plans")
+      mkdirSync(plansDir, { recursive: true })
+
+      const planPath = join(plansDir, "결제-플로우.md")
+      writeFileSync(planPath, "# 결제 플로우\n- [ ] 작업 1")
+
+      const hook = createStartWorkHook(createMockPluginInput())
+      const output = {
+        parts: [
+          {
+            type: "text",
+            text: createStartWorkPrompt({ userRequest: "결제 플로우" }),
+          },
+        ],
+      }
+
+      // when
+      await hook["chat.message"](
+        { sessionID: "session-korean-plan" },
+        output,
+      )
+
+      // then
+      expect(output.parts[0].text).toContain("결제-플로우")
+      expect(output.parts[0].text).toContain("Auto-Selected Plan")
+    })
+
+    test("should match Japanese plan names after Unicode-aware normalization", async () => {
+      // given
+      const plansDir = join(testDir, ".sisyphus", "plans")
+      mkdirSync(plansDir, { recursive: true })
+
+      const planPath = join(plansDir, "支払い-フロー.md")
+      writeFileSync(planPath, "# 支払い フロー\n- [ ] タスク 1")
+
+      const hook = createStartWorkHook(createMockPluginInput())
+      const output = {
+        parts: [
+          {
+            type: "text",
+            text: createStartWorkPrompt({ userRequest: "支払い フロー" }),
+          },
+        ],
+      }
+
+      // when
+      await hook["chat.message"](
+        { sessionID: "session-japanese-plan" },
+        output,
+      )
+
+      // then
+      expect(output.parts[0].text).toContain("支払い-フロー")
+      expect(output.parts[0].text).toContain("Auto-Selected Plan")
+    })
+
+    test("should keep ASCII plan name matching behavior unchanged", async () => {
+      // given
+      const plansDir = join(testDir, ".sisyphus", "plans")
+      mkdirSync(plansDir, { recursive: true })
+
+      const planPath = join(plansDir, "checkout-flow.md")
+      writeFileSync(planPath, "# Checkout Flow\n- [ ] Task 1")
+
+      const hook = createStartWorkHook(createMockPluginInput())
+      const output = {
+        parts: [
+          {
+            type: "text",
+            text: createStartWorkPrompt({ userRequest: "checkout flow" }),
+          },
+        ],
+      }
+
+      // when
+      await hook["chat.message"](
+        { sessionID: "session-ascii-plan" },
+        output,
+      )
+
+      // then
+      expect(output.parts[0].text).toContain("checkout-flow")
+      expect(output.parts[0].text).toContain("Auto-Selected Plan")
+    })
+
+    test("should match mixed ASCII and non-ASCII plan names", async () => {
+      // given
+      const plansDir = join(testDir, ".sisyphus", "plans")
+      mkdirSync(plansDir, { recursive: true })
+
+      const planPath = join(plansDir, "v2-결제-flow.md")
+      writeFileSync(planPath, "# v2 결제 flow\n- [ ] Task 1")
+
+      const hook = createStartWorkHook(createMockPluginInput())
+      const output = {
+        parts: [
+          {
+            type: "text",
+            text: createStartWorkPrompt({ userRequest: "v2 결제 flow" }),
+          },
+        ],
+      }
+
+      // when
+      await hook["chat.message"](
+        { sessionID: "session-mixed-plan" },
+        output,
+      )
+
+      // then
+      expect(output.parts[0].text).toContain("v2-결제-flow")
+      expect(output.parts[0].text).toContain("Auto-Selected Plan")
+    })
   })
 
   describe("session agent management", () => {