thincoder 0.12.54 → 0.12.59

package/src/agent-tools/task.mjs

@@ -31,9 +31,10 @@ export const taskTool = {
   description:
     "Plan and track a task list for complex multi-step work. Each call replaces the entire list. " +
     "Keep exactly one item in_progress at a time; mark items done as you complete them; never mark done if tests fail or work is partial. " +
-    "Statuses: pending | in_progress | done. " +
-    "IMPORTANT: status must be exactly one of these three strings — no synonyms (e.g. 'completed', 'finished', 'open' are INVALID). " +
-    "IMPORTANT: title is required and must be a non-empty string — items with empty titles are silently dropped.",
+    "Statuses: pending | in_progress | done — synonyms (completed/finished/complete, todo/open/waiting, active/running/working, …) are accepted and normalized with a warning. " +
+    "IMPORTANT: title is required and must be a non-empty string — items with empty titles are silently dropped. " +
+    "For cross-session / project-level tracking, use checklist. " +
+    "Returns the updated task list (or the new item's ID on add).",
   parameters: {
     type: "object",
     properties: {

package/src/agent-tools/timer.mjs

@@ -11,25 +11,30 @@ export const timerTool = {
     "Set a timer before you start analyzing code. When the timer fires, " +
     "a system reminder will be injected suggesting you try running code or " +
     "adding debug logs. Use this to enforce a thinking budget: you get " +
-    "N seconds to reason, then the timer reminds you to act.",
+    "N seconds to reason, then the timer reminds you to act. " +
+    "Returns the set confirmation — the reminder fires at the deadline.",
   parameters: {
     type: "object",
     properties: {
       seconds: {
         type: "number",
-        description: "Thinking budget in seconds (default 30). Longer for complex reasoning, shorter for simple tasks.",
+        description: "Thinking budget in seconds (default 180). Longer for complex reasoning, shorter for simple tasks.",
       },
       message: {
         type: "string",
         description: "Custom reminder message to show when time is up. Default: a suggestion to add debug logs or run the code.",
       },
     },
-    required: ["seconds"],
+    required: [],
   },
   readonly: true,
   sideEffectExempt: true,
   execute(args, ctx) {
-    const seconds = args.seconds ?? 180
+    // D15.3#1：默认 180；非法值（非数字/非正数）显式报错——静默 NaN 定时器永不触发的坑
+    const seconds = Number(args.seconds ?? 180)
+    if (!Number.isFinite(seconds) || seconds <= 0) {
+      throw new Error(`timer seconds must be a positive number, got: ${args.seconds}`)
+    }
     const expiresAt = Date.now() + seconds * 1000
     const message = args.message || `⏰ Time's up (${seconds}s). Have you tried running the code, adding a console.log, or checking the output? Thinking more without data is guessing.`
 

package/src/agent-tools/verify.mjs

@@ -1,42 +1,102 @@
 import { isDocFile } from "../advisor/repos.mjs"
 import { execSync, spawn, spawnSync } from "node:child_process"
 import { readFileSync, existsSync } from "node:fs"
-import { join, resolve } from "node:path"
+import { dirname, join, resolve } from "node:path"
 
 /**
- * Source module → test file mapping. Heuristic: the FIRST path component
- * after src/ determines the module. Map it to the test file that imports from it.
- * Modules without dedicated tests map to null.
+ * Source module → test file mapping (AGENT-LOOP §18.14: test files are split by domain —
+ * one module maps to the list of domain files that cover it). Heuristic: the FIRST
+ * path component after src/ determines the module. Modules without dedicated tests map to null.
  */
 const MODULE_TO_TEST = {
-  tools: "test/tools.test.mjs",
-  "agent-tools": "test/tools.test.mjs",
-  agent: "test/agent.test.mjs",
-  memory: "test/memory.test.mjs",
-  tui: "test/tui.test.mjs",
-  provider: "test/integration-provider.mjs",
-  config: "test/integration-provider.mjs",
-  skills: "test/tools.test.mjs",
-  distill: "test/tools.test.mjs",
-  markdown: "test/agent.test.mjs",
-  advisor: "test/advisor.test.mjs",
+  tools: [
+    "test/file-tools.test.mjs", "test/edit-tools.test.mjs", "test/git.test.mjs",
+    "test/checkpoint.test.mjs", "test/bash.test.mjs", "test/ops-scope.test.mjs",
+    "test/execute.test.mjs", "test/checklist.test.mjs", "test/skills-distill.test.mjs",
+    "test/verify-domain.test.mjs", "test/pdf-parse.test.mjs",
+  ],
+  "agent-tools": [
+    "test/subagent-core.test.mjs", "test/subagent-async.test.mjs", "test/subagent-tool.test.mjs",
+    "test/subagent-panel.test.mjs", "test/subagent-scheduler.test.mjs", "test/verify-domain.test.mjs",
+  ],
+  agent: [
+    "test/agent-core.test.mjs", "test/agent-context.test.mjs", "test/guards.test.mjs",
+    "test/eng-delivery.test.mjs", "test/eng-reminders.test.mjs", "test/provider-stream.test.mjs",
+    "test/stream-rules.test.mjs", "test/session-compaction.test.mjs",
+    "test/skills-distill.test.mjs", "test/verify-domain.test.mjs", "test/dispatch.test.mjs",
+  ],
+  memory: ["test/memory.test.mjs"],
+  tui: ["test/tui-render.test.mjs", "test/tui-input.test.mjs", "test/tui-panel.test.mjs", "test/tui-stream.test.mjs", "test/tui-picker.test.mjs", "test/subagent-blocks.test.mjs"],
+  provider: ["test/integration-provider.mjs"],
+  config: ["test/integration-provider.mjs"],
+  skills: ["test/skills-distill.test.mjs"],
+  distill: ["test/skills-distill.test.mjs"],
+  markdown: ["test/memory.test.mjs"],
+  advisor: ["test/advisor-message.test.mjs", "test/advisor-review.test.mjs", "test/advisor-eng.test.mjs", "test/advisor-parallel.test.mjs"],
+  prompts: ["test/prompts.test.mjs"],
+  context: ["test/agent-context.test.mjs", "test/session-compaction.test.mjs"],
+  session: ["test/session.test.mjs", "test/session-compaction.test.mjs", "test/session-eng-advisor.test.mjs", "test/session-migration.test.mjs", "test/session-safety.test.mjs", "test/acp.test.mjs"],
   mcp: null,
-  prompts: null,
-  context: null,
-  session: null,
 }
 
 /**
- * Extract module name from a source path.
- * "src/tools/bash.mjs" → "tools", "src/agent.mjs" → "agent", "src/agent/helpers.mjs" → "agent"
+ * Extract module name from a source path (src-relative or absolute — §18.12:
+ * changed files are normalized to absolute paths before this runs).
+ * "src/tools/bash.mjs" → "tools", "src/agent.mjs" → "agent",
+ * "D:/proj/src/agent/helpers.mjs" → "agent"
  */
 function moduleName(srcPath) {
-  const rel = srcPath.replace(/^src[/\\]/, "")
+  const norm = srcPath.replace(/\\/g, "/")
+  const srcIdx = norm.lastIndexOf("/src/")
+  const rel = srcIdx === -1 ? norm.replace(/^src\//, "") : norm.slice(srcIdx + 5)
   const firstSlash = rel.search(/[/\\]/)
   if (firstSlash === -1) return rel.replace(/\.mjs$/, "")
   return rel.slice(0, firstSlash)
 }
 
+/**
+ * §18.12 D-VR1 path normalization — mirrors the §20.5 file-domain handling:
+ * resolved to an absolute path with forward slashes; dedup uses the win32
+ * lowercase comparison key (changedFileKey) so "D:/x/src/a.mjs" and
+ * "d:\x\SRC\a.mjs" count as the same file.
+ */
+function normalizeChangedPath(p, base) {
+  return resolve(base, p).replace(/\\/g, "/")
+}
+
+/** Win32 comparison key: case-insensitive drives/folders (same file). */
+function changedFileKey(p) {
+  return process.platform === "win32" ? p.toLowerCase() : p
+}
+
+/**
+ * Nearest ancestor of an absolute path holding package.json or .git — the
+ * project/repo root (§18.12 F-VR1: the agent cwd may be a workspace root that
+ * is NOT the repo root — related test files must be located against it).
+ * Walk stops at the filesystem root; returns null when no anchor exists.
+ */
+function findProjectRoot(absPath) {
+  let dir = dirname(resolve(absPath))
+  for (;;) {
+    if (existsSync(join(dir, "package.json")) || existsSync(join(dir, ".git"))) return dir
+    const parent = dirname(dir)
+    if (parent === dir) return null
+    dir = parent
+  }
+}
+
+/**
+ * True when the path is under <projectRoot>/src/ (anchored via findProjectRoot —
+ * a parent-path /src/ segment outside the project must not classify a file as
+ * product code). Falls back to the loose /src/ pattern when no anchor exists.
+ */
+function isUnderSrc(absPath) {
+  const root = findProjectRoot(absPath)
+  if (!root) return /(?:^|[\\/])src[\\/]/.test(absPath)
+  const norm = normalizeChangedPath(absPath, root)
+  return norm.startsWith(normalizeChangedPath("src", root) + "/")
+}
+
 /**
  * verify tool: pre-completion self-check. When called:
  * 0. Doc-only fast path — all changed files are docs (docs/, *.md, LICENSE…):
@@ -52,42 +112,76 @@ function moduleName(srcPath) {
 export const verifyTool = {
   name: "verify",
   description:
-    "Run a pre-completion self-check. By default runs syntax checks on changed files AND any test files related to the changed modules, shows git diff and task list, and displays a self-review checklist. Set full=true to run the project's full test suite (npm test) instead of just related tests. Call this BEFORE declaring any coding task complete — do not say 'done' until verify passes.",
+    "Run a pre-completion self-check. By default runs syntax checks on changed files AND any test files related to the changed modules, shows git diff and task list, and displays a self-review checklist. Set full=true to run the project's full test suite (npm test) instead of just related tests. testNamePattern limits the run to matching test names (renamed from filter — the old name is rejected with an error). Call this BEFORE declaring any coding task complete — do not say 'done' until verify passes. " +
+    "Returns the check report — syntax results, related-test outcomes, task list; verify failure blocks the completion claim.",
   parameters: {
     type: "object",
     properties: {
       full: { type: "boolean", description: "Run the full test suite (npm test) instead of just related tests. Default false — use sparingly, per the testing discipline rules." },
       workdir: { type: "string", description: "Optional: run verify in this subdirectory (relative to cwd or absolute) — for monorepos" },
-      filter: { type: "string", description: "Optional: limit the test run to matching test names (node --test-name-pattern / npm test -- --test-name-pattern)" },
+      testNamePattern: { type: "string", description: "Optional: limit the test run to matching test names (node --test-name-pattern / npm test -- --test-name-pattern). Renamed from filter — the old name is rejected with an error." },
     },
   },
   readonly: true,
   outputPanel: true, // stream test output to a panel instead of inline
   async execute(args, ctx) {
+    if ("filter" in args) {
+      throw new Error("filter was renamed to testNamePattern — use testNamePattern; the old name is rejected")
+    }
     const cwd = ctx.agent.cwd
-    // workdir only relocates WHERE tests (and package.json) live — changed-file
-    // resolution (git diff) stays anchored to the project root.
+    // Changed-file resolution (§18.12 D-VR3): _touchedFiles (per-run bookkeeping,
+    // absolute paths) ∪ git diff fallback — git is tried at testCwd
+    // (workdir-resolved) first, then at ctx.agent.cwd; first success wins.
+    // git-diff paths are repo-root-relative — resolved against the discovered
+    // git root, then normalized (absolute, forward slashes, win32 key).
     const testCwd = args.workdir ? resolve(cwd, args.workdir) : cwd
     const lines = []
     lines.push("=== VERIFICATION REPORT ===")
     lines.push("")
 
-    // 1. Git diff — find changed files
-    let changedFiles = []
-    try {
-      const diff = execSync("git diff --stat", { cwd, encoding: "utf8", stdio: ["ignore", "pipe", "ignore"], timeout: 5000 })
-      if (diff.trim()) {
-        lines.push("Changed files (git diff --stat):")
-        lines.push(diff.trim())
-        // extract changed file paths
-        const nameOnly = execSync("git diff --name-only", { cwd, encoding: "utf8", stdio: ["ignore", "pipe", "ignore"], timeout: 5000 })
-        changedFiles = nameOnly.trim().split("\n").filter(Boolean)
-      } else {
-        lines.push("Changed files: (none — no uncommitted changes)")
-      }
-    } catch {
+    // 1. Changed files — _touchedFiles ∪ git diff (§18.12 D-VR1)
+    let gitOk = false
+    let gitStat = ""
+    let gitFiles = []
+    for (const gitCwd of new Set([testCwd, cwd])) {
+      try {
+        // Anchor the repo root FIRST (also proves git works here) — diff paths
+        // are repo-root-relative, so both diff calls run at the root.
+        const gitRoot = execSync("git rev-parse --show-toplevel", { cwd: gitCwd, encoding: "utf8", stdio: ["ignore", "pipe", "ignore"], timeout: 5000 }).trim() || gitCwd
+        const diff = execSync("git diff --stat", { cwd: gitRoot, encoding: "utf8", stdio: ["ignore", "pipe", "ignore"], timeout: 5000 })
+        const nameOnly = execSync("git diff --name-only", { cwd: gitRoot, encoding: "utf8", stdio: ["ignore", "pipe", "ignore"], timeout: 5000 })
+        gitOk = true
+        gitStat = diff.trim()
+        gitFiles = nameOnly.trim().split("\n").filter(Boolean).map((p) => normalizeChangedPath(p, gitRoot))
+        break
+      } catch { /* git unavailable here — try the next cwd in the chain */ }
+    }
+    if (gitOk && gitStat) {
+      lines.push("Changed files (git diff --stat):")
+      lines.push(gitStat)
+    } else if (gitOk) {
+      lines.push("Changed files: (none — no uncommitted changes)")
+    } else {
       lines.push("Changed files: (not a git repo or git unavailable)")
     }
+    // _touchedFiles: files written by the tools this run — captured regardless
+    // of git state (§18.12 F-VR1: subagent cwd ≠ git root must still locate).
+    const touchedFiles = (ctx.agent._touchedFiles ?? []).map((p) => normalizeChangedPath(p, cwd))
+    const gitKeys = new Set(gitFiles.map(changedFileKey))
+    const touchedOnly = touchedFiles.filter((p) => !gitKeys.has(changedFileKey(p)))
+    if (touchedOnly.length > 0) {
+      lines.push("")
+      lines.push("Changed files (this run — _touchedFiles):")
+      for (const f of touchedOnly) lines.push(`  ${f}`)
+    }
+    // Union, deduped by comparison key (win32 case-insensitive).
+    const changedKeys = new Set()
+    const changedFiles = [...touchedFiles, ...gitFiles].filter((p) => {
+      const k = changedFileKey(p)
+      if (changedKeys.has(k)) return false
+      changedKeys.add(k)
+      return true
+    })
 
     // 1b. Doc-only fast path: every changed file is documentation (docs/, *.md,
     // LICENSE…) — syntax checks and tests are meaningless for doc changes, and
@@ -97,7 +191,7 @@ export const verifyTool = {
     // code — excluded from the fast path, consistent with isProductCode.
     // Empty list (no changes / git unavailable) intentionally falls through
     // to the normal path below.
-    if (changedFiles.length > 0 && changedFiles.every((f) => !/^src[\\/]/.test(f) && isDocFile(f))) {
+    if (changedFiles.length > 0 && changedFiles.every((f) => !isUnderSrc(f) && isDocFile(f))) {
       lines.push("")
       lines.push("Documentation-only changes — skipping syntax checks and tests.")
       ctx.agent._verifyPassed = true
@@ -111,7 +205,7 @@ export const verifyTool = {
       lines.push("")
       lines.push("Syntax check (node --check):")
       for (const f of jsFiles) {
-        const abs = join(cwd, f)
+        const abs = resolve(cwd, f)
         if (!existsSync(abs)) continue // skip deleted files
         try {
           const result = spawnSync("node", ["--check", abs], { encoding: "utf8", stdio: ["ignore", "pipe", "pipe"], timeout: 10000 })
@@ -129,9 +223,9 @@ export const verifyTool = {
     }
 
     // 3. Identify related test files for changed source modules
-    const srcFiles = changedFiles.filter((f) => /^src[/\\].+\.mjs$/i.test(f) && existsSync(join(cwd, f)))
+    const srcFiles = changedFiles.filter((f) => /\.mjs$/i.test(f) && isUnderSrc(f) && existsSync(f))
     const modules = [...new Set(srcFiles.map(moduleName))]
-    const relatedTests = [...new Set(modules.map((m) => MODULE_TO_TEST[m]).filter(Boolean))]
+    const relatedTests = [...new Set(modules.flatMap((m) => MODULE_TO_TEST[m] ?? []).filter(Boolean))]
 
     // 4. Run tests
     const pkgPath = join(testCwd, "package.json")
@@ -142,7 +236,7 @@ export const verifyTool = {
       if (hasTestScript) {
         lines.push("")
         lines.push("Tests (full suite):")
-        const result = await runTestSuite(testCwd, ctx, args.filter)
+        const result = await runTestSuite(testCwd, ctx, args.testNamePattern)
         if (result.passed) {
           lines.push("✓ All tests passed.")
           ctx.agent._verifyPassed = !syntaxFailed
@@ -160,14 +254,28 @@ export const verifyTool = {
       lines.push("")
       lines.push(`Related tests (${relatedTests.length} file(s) for modules: ${modules.join(", ")}):`)
       let anyTestFailed = false
+      let anyTestMissing = false
       for (const testFile of relatedTests) {
-        const abs = join(cwd, testFile)
-        if (!existsSync(abs)) {
-          lines.push(`  ? ${testFile} — file not found, skipping`)
+        // Lookup: cwd first (existing behavior), then the changed files' project
+        // root (§18.12 F-VR1 — subagent cwd may be a workspace root, not the repo
+        // root; the mapped test file lives at <projectRoot>/test/...).
+        let testAbs = join(cwd, testFile)
+        let testRunCwd = cwd
+        if (!existsSync(testAbs)) {
+          const roots = [...new Set(srcFiles.map(findProjectRoot).filter(Boolean))]
+          const inRoot = roots.map((r) => ({ root: r, p: join(r, testFile) })).find((x) => existsSync(x.p))
+          if (inRoot) {
+            testAbs = inRoot.p
+            testRunCwd = inRoot.root
+          }
+        }
+        if (!existsSync(testAbs)) {
+          anyTestMissing = true
+          lines.push(`  ? ${testFile} — file not found (cwd + changed files' project root), verify did not run it`)
           continue
         }
         try {
-          const result = await runTestFile(cwd, testFile, ctx, args.filter)
+          const result = await runTestFile(testRunCwd, testAbs, ctx, args.testNamePattern)
           if (result.passed) {
             lines.push(`  ✓ ${testFile}`)
           } else {
@@ -184,13 +292,17 @@ export const verifyTool = {
         lines.push("")
         lines.push("✗ Related tests FAILED. Review the output above, fix the issues, then run verify again.")
         ctx.agent._verifyPassed = false
+      } else if (anyTestMissing) {
+        lines.push("")
+        lines.push("✗ Related test file(s) MISSING — verify did NOT certify this change. Locate the test file or fix MODULE_TO_TEST.")
+        ctx.agent._verifyPassed = false
       } else {
         lines.push("  All related tests passed.")
         ctx.agent._verifyPassed = !syntaxFailed
       }
     } else {
       // No related tests found for the changed modules
-      const uncovered = modules.filter((m) => MODULE_TO_TEST[m] === null)
+      const uncovered = modules.filter((m) => !MODULE_TO_TEST[m] || (Array.isArray(MODULE_TO_TEST[m]) && MODULE_TO_TEST[m].length === 0))
       const untested = modules.filter((m) => !(m in MODULE_TO_TEST))
       lines.push("")
       if (uncovered.length > 0) {

package/src/agent-tools.mjs

@@ -13,3 +13,4 @@ export { recentChangesTool } from "./agent-tools/recent-changes.mjs"
 export { timerTool } from "./agent-tools/timer.mjs"
 export { advisorTool } from "./agent-tools/advisor.mjs"
 export { engTool } from "./agent-tools/eng.mjs"
+export { readHistoryTool } from "./agent-tools/read-history.mjs"