jules-orchestrator-kit 0.56.0 → 0.57.0

package/README.md

@@ -204,7 +204,7 @@ To maximize PR merge rates, dispatch tasks according to deterministic boundaries
 * **Fail-Closed Security & Secret Redaction:** Evaluates explicit Deny rules before Allow rules against canonicalized, case-folded paths. Redacts high-entropy keys and base64-encoded credentials (such as Kubernetes `Secret` manifests).
 * **Complexity & Cost Router:** Zero-dependency heuristic classifier (`src/router.mjs`) routing mechanical tasks to lightweight models while reserving primary models for complex refactors, with a `node --check` syntax-verification gate that transparently escalates a FAST-tier result to the primary provider if it left broken JS on disk.
 * **Terminal UI & Diagnostic Matrix (`agentctl doctor`):** Interactive terminal dashboard, task sidecar manager, and automated transactional self-repair.
-* **Verified Test Suite:** Tested with **851 unit tests across 115 suites passing in < 14.0s**.
+* **Verified Test Suite:** Tested with **864 unit tests across 121 suites passing in < 15.0s**.
 
 <br/>
 

package/bin/agentctl.mjs

@@ -570,8 +570,16 @@ async function main() {
       if (values.committed) selectedMode = "committed";
       if (values["working-tree"]) selectedMode = "working-tree";
 
-      const minScore = Number(values["min-score"]) || 80;
-      const maxMutants = Number(values["max-mutants"]) || 20;
+      // `Number(x) || default` swallows a legitimate zero: `--min-score 0`,
+      // the way to run the harness for its report without a threshold, silently
+      // enforced 80. Both flags already carry a parseArgs default, so the only
+      // thing left to guard against is a value that is not a number at all.
+      const parseNumericFlag = (raw, fallback) => {
+        const n = Number(raw);
+        return Number.isFinite(n) ? n : fallback;
+      };
+      const minScore = parseNumericFlag(values["min-score"], 80);
+      const maxMutants = parseNumericFlag(values["max-mutants"], 20);
       const testCmd = values.cmd || config.verify?.test || "npm test";
 
       if (!values.json) {
@@ -597,7 +605,11 @@ async function main() {
         console.log(`  • Mutants Killed (Fail)   : ${report.killedMutants} ✅`);
         console.log(`  • Mutants Survived (Pass) : ${report.survivedMutants} ${report.survivedMutants > 0 ? "⚠️" : ""}`);
         console.log(`  • Errors / Timeouts       : ${report.errorMutants}`);
-        console.log(`  • Mutation Score          : ${report.mutationScore}% (Required: ${report.minScore}%)`);
+        console.log(
+          report.scored === false
+            ? `  • Mutation Score          : n/a — ${report.reason}`
+            : `  • Mutation Score          : ${report.mutationScore}% (Required: ${report.minScore}%)`
+        );
         console.log(`  • Duration                : ${report.durationMs}ms\n`);
 
         if (report.survivors.length > 0) {
@@ -1069,7 +1081,29 @@ async function main() {
       });
 
       const queueDir = getQueueDir(root);
-      const files = readdirSync(queueDir).filter((f) => isTaskFile(f, queueDir));
+      // The DAG runner accepts `.json` and `.task` envelopes as well as
+      // Markdown, and does its own discovery — but this count gated whether it
+      // was ever called, using the Markdown-only filter. A queue holding only
+      // JSON envelopes reported "0 queued task(s)" and did nothing, with no
+      // indication that the files were there and understood.
+      const queueEntries = readdirSync(queueDir);
+      let files;
+      if (values.dag) {
+        const { isDagTaskFile } = await import("../src/dag-engine.mjs");
+        files = queueEntries.filter((f) => {
+          if (f === "completed" || f.startsWith(".")) return false;
+          if (!/\.(md|json|task)$/.test(f)) return false;
+          let content = "";
+          try {
+            content = readFileSync(join(queueDir, f), "utf-8");
+          } catch (_) {
+            return false;
+          }
+          return isDagTaskFile(f, content, isTaskFile);
+        });
+      } else {
+        files = queueEntries.filter((f) => isTaskFile(f, queueDir));
+      }
       console.log(`Found ${files.length} queued task(s) in .agent/jules-queue/`);
       if (files.length > 0) {
         const concurrency = values.concurrency ? Number(values.concurrency) : undefined;
@@ -1092,11 +1126,30 @@ async function main() {
     }
 
     case "swarm": {
-      console.log("🚀 Running Swarm Orchestrator...");
+      // This case had no parseArgs at all: `--json` and `--dry-run` were
+      // accepted by the shell, documented in the registry, and silently
+      // discarded — so a rehearsal dispatched for real and a script asking for
+      // JSON got decorated prose.
+      const { values } = parseArgs({
+        args: args.slice(1),
+        options: {
+          concurrency: { type: "string", short: "c" },
+          "dry-run": { type: "boolean", short: "d" },
+          json: { type: "boolean", short: "j" },
+        },
+        allowPositionals: true,
+        strict: false,
+      });
+
       const queueDir = getQueueDir(root);
       const files = readdirSync(queueDir).filter((f) => isTaskFile(f, queueDir));
+      if (!values.json) console.log("🚀 Running Swarm Orchestrator...");
       if (files.length === 0) {
-        console.log("No pending tasks found for swarm.");
+        if (values.json) {
+          console.log(JSON.stringify({ ok: true, processed: 0, results: [], dryRun: Boolean(values["dry-run"]) }, null, 2));
+        } else {
+          console.log("No pending tasks found for swarm.");
+        }
         process.exit(0);
       }
       const tasks = files.map((f) => ({
@@ -1104,7 +1157,16 @@ async function main() {
         title: f.replace(/\.md$/, ""),
         prompt: readFileSync(join(queueDir, f), "utf-8"),
       }));
-      const results = await run(tasks, { root, config, concurrency: config.limits.concurrency || 3 });
+      const results = await run(tasks, {
+        root,
+        config,
+        concurrency: values.concurrency ? Number(values.concurrency) : config.limits.concurrency || 3,
+        dryRun: values["dry-run"],
+      });
+      if (values.json) {
+        console.log(JSON.stringify({ ok: true, dryRun: Boolean(values["dry-run"]), ...results }, null, 2));
+        process.exit((results.results || []).some((r) => r && r.ok === false) ? 1 : 0);
+      }
       process.exit(reportRunOutcome(results));
       break;
     }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "jules-orchestrator-kit",
-  "version": "0.56.0",
+  "version": "0.57.0",
   "description": "Zero-dependency safety gatekeeper, test oracle generator, and multi-agent coordination protocol for autonomous coding agents — Google Jules, Claude Code, Codex and Gemini CLI.",
   "repository": {
     "type": "git",

package/src/assertions.mjs

@@ -513,6 +513,11 @@ export function assertMutation(config = {}, root = process.cwd()) {
   });
 
   const diagnostics = [];
+  if (report.scored === false) {
+    // Not a failure, but not a pass worth trusting either: say so, so a green
+    // stage cannot be read as "the diff survived mutation testing".
+    diagnostics.push(report.reason || "No mutants could be generated from the added lines.");
+  }
   if (!report.ok) {
     diagnostics.push(
       `Mutation score ${report.mutationScore}% below required threshold of ${minScore}% (${report.killedMutants}/${report.totalMutants} killed, ${report.survivedMutants} survived).`

package/src/dag-engine.mjs

@@ -431,7 +431,7 @@ export function resolveAffectedTests(modifiedFiles = [], options = {}) {
  * @param {Function} isTaskFile - `isTaskFile` from engine.mjs, passed in to avoid a circular import.
  * @returns {boolean}
  */
-function isDagTaskFile(fileName, content, isTaskFile) {
+export function isDagTaskFile(fileName, content, isTaskFile) {
   if (fileName.endsWith(".task")) return true;
   if (fileName.endsWith(".md")) return isTaskFile(fileName, content);
   if (fileName.endsWith(".json")) {

package/src/git.mjs

@@ -384,8 +384,17 @@ export function diffText(root = process.cwd(), base = "main", mode = "committed"
         const fullPath = join(root, file);
         if (existsSync(fullPath)) {
           const content = readFileSync(fullPath, "utf-8");
+          const addedLines = content.split(/\r?\n/);
+          // The `@@` header is not decoration. Every consumer that needs to know
+          // *which* line an addition is on parses it — the mutation harness and
+          // the V8 diff-coverage mapper both walk hunks — so a synthetic diff
+          // without one made untracked files invisible to them. Both silently
+          // reported nothing to do for a brand-new file, which is precisely
+          // where untested code arrives. The secret scanner never noticed
+          // because it only reads `+` lines.
           untrackedDiff += `\ndiff --git a/${file} b/${file}\nnew file mode 100644\n--- /dev/null\n+++ b/${file}\n`;
-          untrackedDiff += content.split(/\r?\n/).map((line) => `+${line}`).join("\n") + "\n";
+          untrackedDiff += `@@ -0,0 +1,${addedLines.length} @@\n`;
+          untrackedDiff += addedLines.map((line) => `+${line}`).join("\n") + "\n";
         }
       } catch (_) {}
     }

package/src/mutation.mjs

@@ -676,13 +676,23 @@ export function runMutationTest(options = {}) {
   const selectedMutants = candidates.slice(0, maxMutants);
 
   if (selectedMutants.length === 0) {
+    // A diff with no mutable operators yields no mutants, and reporting that as
+    // "100%" told operators their untested code had a perfect score — the exact
+    // false confidence the harness exists to remove. There is no score to
+    // report, so there is none: `mutationScore` is null and `reason` says why.
+    //
+    // `ok` stays true. Nothing was falsifiable, so nothing failed to be
+    // falsified; failing here would block every diff that only adds imports,
+    // constants or markdown, and a gate that cries wolf gets switched off.
     return {
       ok: true,
       totalMutants: 0,
       killedMutants: 0,
       survivedMutants: 0,
       errorMutants: 0,
-      mutationScore: 100,
+      mutationScore: null,
+      scored: false,
+      reason: "No mutable operators in the added lines — nothing to falsify, so no score was computed.",
       minScore,
       results: [],
       survivors: [],

package/src/ops/command-registry.mjs

@@ -115,23 +115,27 @@ export const COMMAND_REGISTRY = [
   },
   {
     id: "swarm",
+    // Described as an inspector — `mutates: false`, `risk: low` — while the
+    // handler dispatches every queued task in parallel and spends budget.
+    // `--interactive` was advertised and implemented nowhere.
     path: ["swarm"],
     title: "swarm",
-    description: "Inspect active worker slots and concurrency scheduler",
+    description: "Dispatch every queued task in parallel across worker slots",
     category: "Operate",
-    mutates: false,
-    risk: "low",
-    interactive: "optional",
+    mutates: true,
+    risk: "moderate",
+    interactive: "never",
     requiresRepository: true,
     shortcuts: ["s"],
     examples: [
+      "agentctl swarm --dry-run",
       "agentctl swarm",
-      "agentctl swarm --interactive",
-      "agentctl swarm --json",
+      "agentctl swarm --concurrency 3 --json",
     ],
     flags: [
-      { name: "interactive", type: "boolean", description: "Open full-screen swarm dashboard" },
-      { name: "json", type: "boolean", description: "Output structured JSON swarm snapshot" },
+      { name: "concurrency", type: "string", description: "Parallel worker slots (defaults to limits.concurrency)" },
+      { name: "dry-run", type: "boolean", description: "Report what would run without dispatching" },
+      { name: "json", type: "boolean", description: "Output structured JSON swarm result" },
     ],
   },
   {

package/src/ops/tdd-generator.mjs

@@ -1,5 +1,5 @@
 import { writeFileSync, mkdirSync } from "node:fs";
-import { join, basename } from "node:path";
+import { join } from "node:path";
 import { detectPolyglotStack } from "../stack-detector.mjs";
 import { resolveRoot } from "../config.mjs";
 import { runCmd } from "../git.mjs";
@@ -26,45 +26,100 @@ export function scaffoldTddTest(spec = {}, options = {}) {
   const details = spec.spec || "Feature requirement specification assertion.";
 
   const stack = detectPolyglotStack(root);
-  const testDir = join(root, "test");
-  try {
-    mkdirSync(testDir, { recursive: true });
-  } catch (_) {}
 
-  const fileName = `generated-${title}.test.mjs`;
-  const filePath = join(testDir, fileName);
+  // The generated oracle has to be written in the language its runner speaks.
+  //
+  // This used to emit a Node test file for every stack and then, in a Python
+  // project, run `pytest generated-x.test.mjs`. pytest exits 4 on a file it
+  // cannot collect, and the cycle read any non-zero exit as RED — so it
+  // reported a verified failing test, and locked an uncollectable file into
+  // scope.deny, having proven nothing at all.
+  const identifier = title.replace(/-/g, "_");
+
+  // The one string that must appear in the runner's output for the failure to
+  // be the *assertion* failing rather than the file never being collected.
+  const redMarker = `TDD Assertion Failed: Requirement '${title}' is not yet implemented.`;
+  const comment = (prefix) => details.split("\n").map((line) => `${prefix} ${line}`).join("\n");
+
+  let relDir = "test";
+  let fileName = `generated-${title}.test.mjs`;
+  let codeContent;
+  let testCmdParts;
+  let testCmdStr;
+
+  if (stack.stack === "python" || stack.stack === "django") {
+    fileName = `test_generated_${identifier}.py`;
+    codeContent = `def test_tdd_oracle_${identifier}():
+    # REQUIREMENT SPECIFICATION:
+${comment("    #")}
+    is_implemented = False
+    assert is_implemented, "${redMarker}"
+`;
+    testCmdParts = ["pytest", `${relDir}/${fileName}`];
+    testCmdStr = `pytest ${relDir}/${fileName}`;
+  } else if (stack.stack === "cargo") {
+    // Cargo discovers integration tests in tests/, not test/.
+    relDir = "tests";
+    fileName = `generated_${identifier}.rs`;
+    codeContent = `#[test]
+fn tdd_oracle_${identifier}() {
+    // REQUIREMENT SPECIFICATION:
+${comment("    //")}
+    let is_implemented = false;
+    assert!(is_implemented, "${redMarker}");
+}
+`;
+    testCmdParts = ["cargo", "test", "--test", `generated_${identifier}`];
+    testCmdStr = `cargo test --test generated_${identifier}`;
+  } else if (stack.stack === "go") {
+    fileName = `generated_${identifier}_test.go`;
+    codeContent = `package ${relDir}
 
-  const codeContent = `import test from "node:test";
+import "testing"
+
+func TestTddOracle${identifier.replace(/_/g, "")}(t *testing.T) {
+	// REQUIREMENT SPECIFICATION:
+${comment("\t//")}
+	isImplemented := false
+	if !isImplemented {
+		t.Fatalf("${redMarker}")
+	}
+}
+`;
+    testCmdParts = ["go", "test", `./${relDir}/`];
+    testCmdStr = `go test ./${relDir}/`;
+  } else {
+    codeContent = `import test from "node:test";
 import assert from "node:assert/strict";
 
 test("TDD Oracle: ${title}", () => {
   // REQUIREMENT SPECIFICATION:
-  // ${details.replace(/\n/g, "\n  // ")}
+${comment("  //")}
 
   const isImplemented = false;
-  assert.equal(isImplemented, true, "TDD Assertion Failed: Requirement '${title}' is not yet implemented.");
+  assert.equal(isImplemented, true, "${redMarker}");
 });
 `;
+    testCmdParts = ["node", "--test", join(root, relDir, fileName)];
+    testCmdStr = `node --test ${relDir}/${fileName}`;
+  }
 
-  writeFileSync(filePath, codeContent, "utf-8");
+  const testDir = join(root, relDir);
+  try {
+    mkdirSync(testDir, { recursive: true });
+  } catch (_) {}
 
-  const relativePath = `test/${fileName}`;
-  let testCmd = ["node", "--test", filePath];
-  let testCmdStr = `node --test ${relativePath}`;
-  if (stack.stack === "python") {
-    testCmd = ["pytest", filePath];
-    testCmdStr = `pytest ${relativePath}`;
-  } else if (stack.stack === "cargo") {
-    testCmd = ["cargo", "test", "--test", title];
-    testCmdStr = `cargo test --test ${title}`;
-  }
+  const filePath = join(testDir, fileName);
+  writeFileSync(filePath, codeContent, "utf-8");
 
   return {
     filePath,
-    relativePath,
+    relativePath: `${relDir}/${fileName}`,
     codeContent,
-    testCmd,
+    testCmd: testCmdParts,
     testCmdStr,
+    stack: stack.stack,
+    redMarker,
   };
 }
 
@@ -98,7 +153,21 @@ export async function runTddCycle(spec = {}, options = {}) {
 
   if (redOutput.status === 0) {
     throw new TddError(
-      `TDD RED check failed: Test 'test/${basename(scaffolded.filePath)}' passed initially! TDD tests must be falsifiable and fail before implementation.`
+      `TDD RED check failed: Test '${scaffolded.relativePath}' passed initially! TDD tests must be falsifiable and fail before implementation.`
+    );
+  }
+
+  // A non-zero exit is not proof the assertion ran. `pytest` exits 4 on a file
+  // it cannot collect and 5 when it collects nothing; a runner pointed at a
+  // file it does not understand exits non-zero for that reason alone. Reading
+  // either as RED is how a Node test file in a Python project came to be
+  // reported as a verified failing oracle. The generated assertion carries a
+  // marker; if the runner never reached it, the marker is not in the output.
+  const redText = `${redOutput.stdout || ""}\n${redOutput.stderr || ""}`;
+  if (scaffolded.redMarker && !redText.includes(scaffolded.redMarker)) {
+    throw new TddError(
+      `TDD RED check inconclusive: '${scaffolded.testCmdStr}' exited ${redOutput.status} without reaching the generated assertion, so nothing was proven falsifiable. ` +
+        `The runner most likely could not collect '${scaffolded.relativePath}' (detected stack: ${scaffolded.stack}). Output:\n${redText.trim().slice(0, 500)}`
     );
   }