agent-roast 0.21.0 → 0.23.0

package/README.md

@@ -37,7 +37,7 @@ npx agent-roast          # review what you have not committed yet (on a clean tr
 ## Stop your coding agent from doing it
 
 ```bash
-npx agent-roast hook install          # Claude Code, Codex and Cursor
+npx agent-roast hook install          # Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot and Factory Droid
 ```
 
 For Claude Code there is also a plugin: `/plugin install agent-roast --marketplace Nari-Ab/agent-roast`. Any agent that speaks
@@ -112,7 +112,7 @@ Measured, with the denominators and the caveats in [docs/benchmarks.md](docs/ben
 deliberate, and not specific to AI. agent-roast is a tripwire that asks about them. On repositories it had not seen, between a quarter and
 two thirds of its blocks were questions with a harmless answer, depending on the sample.**
 
-- **162 of 162** hand-labeled cheating edits are caught and **150 of 150** harmless look-alikes (reformats, moves, raised
+- **164 of 164** hand-labeled cheating edits are caught and **153 of 153** harmless look-alikes (reformats, moves, raised
   thresholds, environment-gated skips) stay silent. The cases were written with the detector, so they guard against regressions;
   they do not prove accuracy. One gap is documented and still open.
 - **A catalog of cheats that someone else wrote.** Of the nine hacks in [reward-hack-zoo](https://github.com/MatthewWaller/reward-hack-zoo) (hand-written specimens; none from a model) four are reported (a skip, a patched target, a `conftest.py` that rewrites reports,
@@ -122,8 +122,8 @@ two thirds of its blocks were questions with a harmless answer, depending on the
   (Claude, Copilot, Cursor, Codex, Devin, aider) and on 0.3% of the others. Inside the same repositories the odds for an
   AI-signed commit were 0.81 times those of another commit (95% CI 0.40–1.67). There is no sign that AI-signed commits weaken
   checks more often.
-- **On repositories it was not written from, between a quarter and two thirds of the blocked commits were false alarms.** Nine samples of 39
-  or 40 public repositories (356 in all, 102,210 commits) were replayed and every blocked commit was read. The share was 69% for 0.17.0 on the
+- **On repositories it was not written from, between a quarter and two thirds of the blocked commits were false alarms.** Ten samples of 39
+  or 40 public repositories (396 in all, 113,401 commits) were replayed and every blocked commit was read. The share was 69% for 0.17.0 on the
   first sample and on the fourth, 51% for 0.18.0 on the fourth and 27% on the fifth, which it had not seen either. This version, frozen before the
   sixth sample was drawn, blocked 100 of its 11,335 commits: 34 changed a check a reviewer should see (an assertion that got weaker, a gate
   that was relaxed, a test switched off or left focused) and 66 were false alarms, but 51 of the 66 were one defect in one repository, an array of
@@ -131,8 +131,8 @@ two thirds of its blocks were questions with a harmless answer, depending on the
   test that asserts nothing counts as a block for a reviewer; counted as false alarms, 13 of 33 on the fifth sample and 72 of 100 on the sixth. Most of the rest, in the
   earlier measurements too, were deliberate and explained. One reader (an AI assistant) sorted them, and the fixes were made from the same
   sorting, so treat the figures as a first estimate. The published 0.19.0 blocked 31 of the 12,000 commits of the seventh sample: 16 for a reviewer and 15 false alarms (48%, 6 of them one
-  repository's lint configuration); the build with the fixes from reading it, tried on an eighth sample that it had not seen, blocked 26 of 11,081 commits: 16 for a reviewer and 10 false alarms (38%), and the fixes from reading that one are in this release too; this release as it was frozen blocked 46 of 11,286 commits of a ninth sample that it had not seen: 28 for a reviewer and 18 false alarms (39%), and five of those are fixed in it ([section 9](docs/benchmarks.md#9-repositories-nobody-had-tuned-on),
-  [section 10](docs/benchmarks.md#10-two-more-samples-and-what-the-fixes-from-sample-4-did), [section 11](docs/benchmarks.md#11-a-seventh-sample-and-two-things-that-were-missing), [section 12](docs/benchmarks.md#12-an-eighth-sample-the-first-measurement-of-the-fixes-from-sample-7), [section 14](docs/benchmarks.md#14-a-ninth-sample-the-shipped-build-on-repositories-it-had-not-seen)).
+  repository's lint configuration); the build with the fixes from reading it, tried on an eighth sample that it had not seen, blocked 26 of 11,081 commits: 16 for a reviewer and 10 false alarms (38%), and the fixes from reading that one are in this release too; 0.21.0 as it was frozen blocked 46 of 11,286 commits of a ninth sample that it had not seen: 28 for a reviewer and 18 false alarms (39%), and five of those are fixed in it; 0.21.0 as it shipped blocked 40 of the 11,191 commits of a tenth sample that it had not seen either: 27 for a reviewer and 13 false alarms (33%), and three of those 13 are fixed in this release ([section 9](docs/benchmarks.md#9-repositories-nobody-had-tuned-on),
+  [section 10](docs/benchmarks.md#10-two-more-samples-and-what-the-fixes-from-sample-4-did), [section 11](docs/benchmarks.md#11-a-seventh-sample-and-two-things-that-were-missing), [section 12](docs/benchmarks.md#12-an-eighth-sample-the-first-measurement-of-the-fixes-from-sample-7), [section 14](docs/benchmarks.md#14-a-ninth-sample-the-shipped-build-on-repositories-it-had-not-seen), [section 15](docs/benchmarks.md#15-a-tenth-sample-the-build-that-ships-on-repositories-it-had-not-seen), [section 16](docs/benchmarks.md#16-what-sample-10-led-to-three-fixes-fitted-and-a-guard-that-did-not-pay)).
 - **9,732 commits compared in full** from 164 public repositories that use coding agents (5,454 of them AI-signed) were reviewed again for the
   last two releases, and every new finding was read. 0.15.0's rules reported 30 new findings there and 7 were false alarms (fixed in 0.16.0);
   0.17.0's rule for an exact check turned into a bound reported 18 in 14 commits, all of them loosened checks, after two false alarms were

package/dist/{chunk-XVV5FNDV.js → chunk-L4XLCXC2.js}

@@ -1508,6 +1508,7 @@ function isCodeFile(filePath, customIgnorePaths) {
 function isTestFile(filePath) {
   return TEST_PATH_REGEX.test(filePath);
 }
+var PYTHON_TEST_PATH = /(?:^|[/._-])(?:tests?|testing|specs?|conftest)(?:[/._-]|$)/i;
 function auditFileDiffs(files, commitHash = "HEAD", commitDate = (/* @__PURE__ */ new Date()).toISOString(), options = {}) {
   const infractions = [];
   let hasSupportedChanges = false;
@@ -1538,7 +1539,7 @@ function auditFileDiffs(files, commitHash = "HEAD", commitDate = (/* @__PURE__ *
       for (const start of masked.templateStarts) testCode[start] = "
";
       maskedLines.push({ code: cleanedLine, testCode: testCode.join(""), raw: rawLine, number: added.lineNumber });
       const isComment = !cleanedLine.trim();
-      if (python2 && !isComment) {
+      if (python2 && !isComment && PYTHON_TEST_PATH.test(file.path)) {
         for (const rule of PYTHON_TEST_SKIP_PATTERNS) {
           if (rule.regex.test(cleanedLine)) {
             if (rule.imperative && isGuardedSkip(maskedLines)) break;
@@ -1644,7 +1645,7 @@ function auditFileDiffs(files, commitHash = "HEAD", commitDate = (/* @__PURE__ *
     }
     if (python2) {
       const reportedSkips = new Set(infractions.slice(firstFinding).filter((f) => f.type === "test-skip").map((f) => f.lineNumber));
-      for (const match of source.matchAll(PYTHON_MARKER)) {
+      for (const match of PYTHON_TEST_PATH.test(file.path) ? source.matchAll(PYTHON_MARKER) : []) {
         const xfail = match[2] === "xfail";
         let open2 = match.index + match[0].length;
         while (source[open2] === " " || source[open2] === "	") open2++;
@@ -4407,11 +4408,13 @@ var THRESHOLD_PATTERNS = [
 var SWITCHES = ["strict", "noImplicitAny", "strictNullChecks", "noImplicitReturns", "useUnknownInCatchVariables"];
 var GATE_SCRIPT = /^(?:pre|post)?(?:test|lint|typecheck|type-check|check|ci|verify|validate)(?::[\w:.-]+)?$/;
 var CI_RUNNER = new RegExp(
-  String.raw`(?<![\w./-])(?:pytest|vitest|jest|mocha|tox|nox)(?![\w/-])` + String.raw`|(?<![\w./-])(?:npm|yarn|pnpm|bun|go|cargo|mvn|gradle|make|dotnet|flutter)\s+(?:run\s+)?(?:-\S+\s+)*test(?![\w/])` + // A package script that runs the tests with coverage: `yarn coverage`, `npm run coverage`.
-  String.raw`|(?<![\w./-])(?:npm|yarn|pnpm|bun)\s+(?:run\s+)?coverage(?![\w/-])` + // Monorepo runners: `nx affected --target=test`, `turbo run test`, `lerna run test`, `bazel test //...`.
+  String.raw`(?<![\w./-])(?:pytest|vitest|jest|mocha|tox|nox)(?![\w/-])` + String.raw`|(?<![\w./-])(?:npm|yarn|pnpm|bun|go|cargo|mvn|gradle|make|dotnet|flutter)\s+(?:run\s+)?(?:-\S+\s+)*test(?![\w/])` + // A package script that runs the tests with coverage: `yarn coverage`, `npm run coverage`, `npm run cover`.
+  String.raw`|(?<![\w./-])(?:npm|yarn|pnpm|bun)\s+(?:run\s+)?cover(?:age)?(?![\w/-])` + // Monorepo runners: `nx affected --target=test`, `turbo run test`, `lerna run test`, `bazel test //...`.
   String.raw`|(?<![\w./-])(?:nx|turbo|lerna|rush)\s+(?:run-many|affected|run|test)\b[^\n]*(?<![\w-])(?:test|tests)(?![\w-])` + String.raw`|(?<![\w./-])(?:bazel\s+test|ctest|phpunit|rspec|deno\s+test)(?![\w/-])` + // Python's own runners and the browser runners: `python3 -m unittest`, `nosetests`, `playwright test`, `cypress run`, `karma start`.
   String.raw`|(?<![\w./-])python[\d.]*\s+-m\s+(?:unittest|pytest|nose2?)(?![\w/-])` + String.raw`|(?<![\w./-])(?:nosetests|nose2|py\.test|stestr|behave|jasmine|playwright\s+test|cypress\s+run|karma\s+start|hatch\s+(?:run\s+)?test|swift\s+test)(?![\w/-])` + // Task runners: `deno task test`, `just test`, `rake test`, `composer test`, `mix test`, `sbt test`, `./gradlew test`, and Node's own `node --test`.
-  String.raw`|(?<![\w./-])(?:deno\s+(?:task|run)|just|task|poe|rake|composer(?:\s+run)?|mix|sbt|lein|stack|cabal|(?:\./)?(?:gradlew|mvnw)(?:\.bat|\.cmd)?)\s+(?:-\S+\s+)*test(?![\w/-])` + String.raw`|(?<![\w./-])(?:node|tsx)(?:\s+-\S+)*\s+--test(?![\w-])` + // A project's own test script, run through its interpreter or by path: `python scripts/run_tests.py`, `./scripts/test.sh`, `python tests/test_cli.py`.
+  String.raw`|(?<![\w./-])(?:deno\s+(?:task|run)|just|task|poe|rake|composer(?:\s+run)?|mix|sbt|lein|stack|cabal|(?:\./)?(?:gradlew|mvnw)(?:\.bat|\.cmd)?)\s+(?:-\S+\s+)*test(?![\w/-])` + String.raw`|(?<![\w./-])(?:node|tsx)(?:\s+-\S+)*\s+--test(?![\w-])` + // A target of a task runner that has the word as one of its own words and not as the first: `make unit-test`, `just test-asyncio py314`, `pnpm run docs:test`. A target that sets up, installs, builds or type-checks
+  // (`make setup-tests`, `pnpm run typecheck:tests`, `pnpm run build:native:test`) runs no tests.
+  String.raw`|(?<![\w./-])(?:make|just|task|(?:npm|pnpm|bun)\s+run|yarn(?:\s+run)?|pnpm)\s+(?:-\S+\s+)*(?![\w.:-]*(?:setup|teardown|install|deps|typecheck|lint|build|docker|clean|prepare|publish|deploy|release|format|image|download|data))[\w.:-]*(?<![A-Za-z0-9])tests?(?![A-Za-z0-9])` + // A project's own test script, run through its interpreter or by path: `python scripts/run_tests.py`, `./scripts/test.sh`, `python tests/test_cli.py`.
   String.raw`|(?<![\w./-])(?:(?:python[\d.]*|node|bash|sh|zsh|ruby|pwsh|tsx|ts-node)(?:\s+-\S+)*\s+|\./)(?:[\w.-]+/)*(?:(?:run[_-]?)?tests?(?:[_-](?:all|suite|runner|unit|ci))?|test_[\w-]+|[\w-]+_test)\.(?:py|sh|bash|js|mjs|cjs|ts|rb|ps1)(?![\w-])` + String.raw`|\brun:.*(?<=\s)(?:tests?|unit|e2e)(?![\w/.-])`
 );
 var SHELL_TEST = /(?:^|run:\s*|[;&|({]\s*|\b(?:if|then|do|else|elif|while|until)\s+|!\s+)test\s+!?\s*(?:-[bcdefghkLnOGprsStuwxzN](?=\s)|["$'])/g;
@@ -4531,12 +4534,16 @@ function ciTestSteps(file, oldText, newText) {
   const after = commandsAfter.length;
   return { file, before: commandsBefore.length, after, reusableAfter: matchingLines(now, CI_REUSABLE), ungatedAfter: after === 0 ? ciTestLines(now).length : 0, commandsBefore, commandsAfter };
 }
-function resolveCiTestSteps(counts, commit, elsewhere = /* @__PURE__ */ new Set()) {
+function gatingCalls(file, text) {
+  if (!(/\.gitlab-ci/.test(file) || CI_GATE.test(gatingTriggers(text) ?? text))) return [];
+  return [...text.matchAll(/^\s*-?\s*uses:\s*["']?\.\/(\.github\/workflows\/[^\s"'#@]+\.ya?ml)/gm)].map((match) => match[1]);
+}
+function resolveCiTestSteps(counts, commit, elsewhere = /* @__PURE__ */ new Set(), called = /* @__PURE__ */ new Set()) {
   const before = counts.reduce((sum, item) => sum + item.before, 0);
   const after = counts.reduce((sum, item) => sum + item.after + item.reusableAfter, 0);
   if (after >= before) return [];
   const running = /* @__PURE__ */ new Set([...counts.flatMap((item) => item.commandsAfter ?? []), ...elsewhere]);
-  return counts.filter((item) => item.before > 0 && item.after === 0 && item.reusableAfter === 0).filter((item) => item.commandsBefore === void 0 || item.commandsBefore.some((command) => !running.has(command))).map((item) => item.ungatedAfter > 0 ? finding("guard-weakened", item.file, 1, "tests no longer run on push or pull_request", "Workflow keeps its tests but no longer runs on pushes or pull requests", commit) : finding("guard-weakened", item.file, 1, "test command no longer runs in CI", "Test step removed from CI", commit));
+  return counts.filter((item) => item.before > 0 && item.after === 0 && item.reusableAfter === 0).filter((item) => !(item.ungatedAfter > 0 && called.has(item.file))).filter((item) => item.commandsBefore === void 0 || item.commandsBefore.some((command) => !running.has(command))).map((item) => item.ungatedAfter > 0 ? finding("guard-weakened", item.file, 1, "tests no longer run on push or pull_request", "Workflow keeps its tests but no longer runs on pushes or pull requests", commit) : finding("guard-weakened", item.file, 1, "test command no longer runs in CI", "Test step removed from CI", commit));
 }
 var LENIENT = /continue-on-error:\s*true|allow_failure:\s*true|^\s*(?:-\s+)?if:\s*(?:\$\{\{\s*)?false\s*(?:\}\})?\s*(?:#.*)?$/;
 var SHELL_SWALLOW = /\|\|\s*(?:true|exit\s+0|:|echo\b)|;\s*(?:true|exit\s+0)\b/;
@@ -4792,6 +4799,7 @@ async function compareSnapshot(options, cwd, snapshot) {
   const tamperingFiles = [];
   const guardFindings = [];
   const ciCounts = [];
+  const calledInChange = /* @__PURE__ */ new Set();
   let productionChanged = false;
   const decode = (content) => {
     totalBytes += content.length;
@@ -4830,6 +4838,7 @@ async function compareSnapshot(options, cwd, snapshot) {
         guardFindings.push(...analyzeGuardFile(change.newPath, oldText, newText, refs.headSha));
         const ci = ciTestSteps(change.newPath, oldText, newText);
         if (ci) ciCounts.push(ci);
+        if (ci && newText !== null) for (const workflow of gatingCalls(change.newPath, newText)) calledInChange.add(workflow);
         result.filesInspected++;
       } catch (error) {
         assertNotAborted(options.signal);
@@ -4887,14 +4896,16 @@ async function compareSnapshot(options, cwd, snapshot) {
   await snapshot?.verify();
   result.complete = result.skippedFiles.length === 0;
   const rules = policy.config.rules ?? {};
-  const elsewhere = /* @__PURE__ */ new Set();
+  const elsewhere = /* @__PURE__ */ new Set(), called = new Set(calledInChange);
   if (ciCounts.some((item) => item.before > 0 && item.after === 0 && item.reusableAfter === 0)) {
     const touched = new Set(changes.flatMap((change) => [change.oldPath, change.newPath]));
     const tree = /^[a-f0-9]{40,64}$/.test(refs.headSha) ? refs.headSha : refs.baseSha;
     try {
       for (const file of (await ciFilesAt(cwd, tree, options.signal)).filter((candidate) => !touched.has(candidate.path)).slice(0, 100)) {
         try {
-          for (const command of enforcedTestCommands(file.path, new TextDecoder().decode(await readBlob(cwd, file.oid, options.signal, 512 * 1024)))) elsewhere.add(command);
+          const text = new TextDecoder().decode(await readBlob(cwd, file.oid, options.signal, 512 * 1024));
+          for (const command of enforcedTestCommands(file.path, text)) elsewhere.add(command);
+          for (const workflow of gatingCalls(file.path, text)) called.add(workflow);
         } catch (error) {
           assertNotAborted(options.signal);
         }
@@ -4903,7 +4914,7 @@ async function compareSnapshot(options, cwd, snapshot) {
       assertNotAborted(options.signal);
     }
   }
-  result.infractions.push(...[...resolveTampering(tamperingFiles, productionChanged), ...guardFindings, ...resolveCiTestSteps(ciCounts, refs.headSha, elsewhere)].filter((finding2) => rules[finding2.type] !== "off"));
+  result.infractions.push(...[...resolveTampering(tamperingFiles, productionChanged), ...guardFindings, ...resolveCiTestSteps(ciCounts, refs.headSha, elsewhere, called)].filter((finding2) => rules[finding2.type] !== "off"));
   const exceptions = applyExceptions(result.infractions, options.ignoreExceptions ? [] : policy.config.exceptions, /* @__PURE__ */ new Date(), existingCounts);
   result.infractions = exceptions.findings;
   result.suppressedFindings = exceptions.suppressedFindings;
@@ -5041,13 +5052,14 @@ function formatReplayTerminal(result, allRules = false) {
 // src/hook.ts
 import path8 from "path";
 import * as fs7 from "fs/promises";
-var HOOK_AGENTS = ["claude", "codex", "cursor"];
+var HOOK_AGENTS = ["claude", "codex", "cursor", "gemini", "copilot", "droid"];
 var DEFAULT_TIMEOUT_MS = 25e3;
 var PAUSE_FILE = "agent-roast-hook-paused";
 var LISTED = 6;
 var LISTED_WARNINGS = 3;
 function detectAgent(input) {
   if ("workspace_roots" in input || "cursor_version" in input || "conversation_id" in input) return "cursor";
+  if (input.hook_event_name === "AfterAgent" || "prompt_response" in input) return "gemini";
   if ("turn_id" in input) return "codex";
   return "claude";
 }
@@ -5055,7 +5067,7 @@ function inputCwd(agent, input, env) {
   const roots = input.workspace_roots;
   if (agent === "cursor" && Array.isArray(roots) && typeof roots[0] === "string") return roots[0];
   if (typeof input.cwd === "string" && input.cwd) return input.cwd;
-  return env.CLAUDE_PROJECT_DIR || env.CURSOR_PROJECT_DIR || process.cwd();
+  return env.CLAUDE_PROJECT_DIR || env.CURSOR_PROJECT_DIR || env.GEMINI_PROJECT_DIR || env.FACTORY_PROJECT_DIR || process.cwd();
 }
 function clip(text, limit) {
   const flat = terminalText(text).replace(/\s+/g, " ").trim();
@@ -5133,8 +5145,8 @@ async function runHook(rawInput, options = {}) {
   if (["off", "0", "false"].includes(mode)) return outcome("skip");
   const advisory = options.advisory === true || mode === "advisory";
   const agent = options.agent ?? detectAgent(input);
-  const event = String(input.hook_event_name ?? "Stop");
-  if (event.toLowerCase() !== "stop") return outcome("skip");
+  const finishing = agent === "gemini" ? "afteragent" : "stop";
+  if (String(input.hook_event_name ?? finishing).toLowerCase() !== finishing) return outcome("skip");
   if (agent === "cursor" && typeof input.status === "string" && input.status !== "completed") return outcome("skip");
   const cwd = options.cwd ?? inputCwd(agent, input, env);
   if (await hookPausedUntil(cwd, options.now)) return outcome("skip");
@@ -5150,15 +5162,17 @@ async function runHook(rawInput, options = {}) {
   } finally {
     clearTimeout(timer);
   }
-  if (!result.failedRules.length) return outcome("pass", agent === "cursor" ? "{}\n" : "", "", result);
+  if (!result.failedRules.length) return outcome("pass", agent === "cursor" || agent === "gemini" ? "{}\n" : "", "", result);
   const alreadyAsked = input.stop_hook_active === true || typeof input.loop_count === "number" && input.loop_count > 0;
   if (alreadyAsked || advisory) {
     const warning = formatUserWarning(result);
-    return outcome("warn", agent === "cursor" ? "{}\n" : JSON.stringify({ systemMessage: warning }) + "\n", agent === "cursor" && advisory ? `${warning}
+    const notice = agent === "copilot" ? { type: "progress", message: warning } : { systemMessage: warning };
+    return outcome("warn", agent === "cursor" ? "{}\n" : JSON.stringify(notice) + "\n", agent === "cursor" && advisory ? `${warning}
 ` : "", result);
   }
   const message = formatAgentFeedback(result);
-  return outcome("block", JSON.stringify(agent === "cursor" ? { followup_message: message } : { decision: "block", reason: message }) + "\n", "", result);
+  const answer = agent === "cursor" ? { followup_message: message } : { decision: agent === "gemini" ? "deny" : "block", reason: message };
+  return outcome("block", JSON.stringify(answer) + "\n", "", result);
 }
 
 // src/hook-install.ts
@@ -5168,8 +5182,9 @@ import path9 from "path";
 import { randomUUID as randomUUID4 } from "crypto";
 import { createRequire as createRequire2 } from "module";
 import { spawnSync as spawnSync3 } from "child_process";
-var AGENT_LABELS = { claude: "Claude Code", codex: "Codex", cursor: "Cursor" };
+var AGENT_LABELS = { claude: "Claude Code", codex: "Codex", cursor: "Cursor", gemini: "Gemini CLI", copilot: "GitHub Copilot", droid: "Factory Droid" };
 var TIMEOUT_SECONDS = 60;
+var FINISH_EVENT = { claude: "Stop", codex: "Stop", cursor: "stop", gemini: "AfterAgent", copilot: "Stop", droid: "Stop" };
 function packageVersion() {
   const require3 = createRequire2(import.meta.url);
   return String(require3("../package.json").version);
@@ -5196,6 +5211,9 @@ function hookConfigPath(agent, scope, root, home, env = process.env) {
   const claude = path9.join(scope === "global" ? home : root, ".claude");
   if (agent === "claude") return path9.join(claude, scope === "local" ? "settings.local.json" : "settings.json");
   if (agent === "cursor") return path9.join(scope === "global" ? home : root, ".cursor", "hooks.json");
+  if (agent === "gemini") return path9.join(scope === "global" ? home : root, ".gemini", "settings.json");
+  if (agent === "copilot") return path9.join(scope === "global" ? env.COPILOT_HOME || path9.join(home, ".copilot") : path9.join(root, ".github"), "hooks", "agent-roast.json");
+  if (agent === "droid") return path9.join(scope === "global" ? home : root, ".factory", "hooks.json");
   return path9.join(scope === "global" ? env.CODEX_HOME || path9.join(home, ".codex") : path9.join(root, ".codex"), "hooks.json");
 }
 function readJsonFile(file) {
@@ -5231,40 +5249,47 @@ function writeJsonFile(link, json) {
   }
 }
 var asArray = (value) => Array.isArray(value) ? value : [];
+var isObject = (value) => Boolean(value) && typeof value === "object" && !Array.isArray(value);
+function eventMap(agent, data, create) {
+  if (agent === "droid") return data;
+  if (create && !isObject(data.hooks)) data.hooks = {};
+  return isObject(data.hooks) ? data.hooks : void 0;
+}
+var isFlat = (agent) => agent === "cursor" || agent === "copilot";
 function addStopHook(agent, data, command) {
   const before = JSON.stringify(data);
-  data.hooks = data.hooks && typeof data.hooks === "object" && !Array.isArray(data.hooks) ? data.hooks : {};
-  if (agent === "cursor") {
+  const events = eventMap(agent, data, true), event = FINISH_EVENT[agent];
+  if (isFlat(agent)) {
     data.version ??= 1;
-    const handlers = asArray(data.hooks.stop);
+    const handlers = asArray(events[event]);
     const ours = handlers.find((handler) => isOurs(handler?.command));
     if (ours) ours.command = command;
-    else handlers.push({ command, timeout: TIMEOUT_SECONDS, loop_limit: 2 });
-    data.hooks.stop = handlers;
+    else handlers.push(agent === "cursor" ? { command, timeout: TIMEOUT_SECONDS, loop_limit: 2 } : { type: "command", command, timeoutSec: TIMEOUT_SECONDS });
+    events[event] = handlers;
   } else {
-    const groups = asArray(data.hooks.Stop);
+    const groups = asArray(events[event]);
     const ours = groups.flatMap((group) => asArray(group?.hooks)).find((handler) => isOurs(handler?.command));
     if (ours) ours.command = command;
-    else groups.push({ hooks: [{ type: "command", command, timeout: TIMEOUT_SECONDS }] });
-    data.hooks.Stop = groups;
+    else groups.push(agent === "gemini" ? { matcher: "*", hooks: [{ name: "agent-roast", type: "command", command, timeout: TIMEOUT_SECONDS * 1e3 }] } : { hooks: [{ type: "command", command, timeout: TIMEOUT_SECONDS }] });
+    events[event] = groups;
   }
   return JSON.stringify(data) !== before;
 }
 function removeStopHook(agent, data) {
   const before = JSON.stringify(data);
-  const hooks = data.hooks;
-  if (!hooks || typeof hooks !== "object") return false;
-  if (agent === "cursor") {
-    const kept = asArray(hooks.stop).filter((handler) => !isOurs(handler?.command));
-    if (kept.length) hooks.stop = kept;
-    else delete hooks.stop;
+  const events = eventMap(agent, data, false), event = FINISH_EVENT[agent];
+  if (!events) return false;
+  if (isFlat(agent)) {
+    const kept = asArray(events[event]).filter((handler) => !isOurs(handler?.command));
+    if (kept.length) events[event] = kept;
+    else delete events[event];
   } else {
-    const groups = asArray(hooks.Stop).map((group) => ({ ...group, hooks: asArray(group?.hooks).filter((handler) => !isOurs(handler?.command)) }));
+    const groups = asArray(events[event]).map((group) => ({ ...group, hooks: asArray(group?.hooks).filter((handler) => !isOurs(handler?.command)) }));
     const kept = groups.filter((group) => group.hooks.length);
-    if (kept.length) hooks.Stop = kept;
-    else delete hooks.Stop;
+    if (kept.length) events[event] = kept;
+    else delete events[event];
   }
-  if (!Object.keys(hooks).length) delete data.hooks;
+  if (events !== data && !Object.keys(events).length) delete data.hooks;
   return JSON.stringify(data) !== before;
 }
 function context(options) {
@@ -5294,14 +5319,15 @@ function uninstallHooks(options) {
 function hookStatus(options) {
   const { root, home, env } = context({ ...options, scope: "project" });
   const rows = [];
-  for (const agent of ["claude", "codex", "cursor"]) {
+  for (const agent of ["claude", "codex", "cursor", "gemini", "copilot", "droid"]) {
     for (const scope of ["project", "local", "global"]) {
       if (scope === "local" && agent !== "claude") continue;
       const file = hookConfigPath(agent, scope, root, home, env);
       let command;
       try {
         const data = readJsonFile(file).data;
-        const handlers = agent === "cursor" ? asArray(data.hooks?.stop) : asArray(data.hooks?.Stop).flatMap((group) => asArray(group?.hooks));
+        const listed = asArray(eventMap(agent, data, false)?.[FINISH_EVENT[agent]]);
+        const handlers = isFlat(agent) ? listed : listed.flatMap((group) => asArray(group?.hooks));
         command = handlers.map((handler) => handler?.command).find(isOurs);
       } catch {
       }

package/dist/cli.js

@@ -38,7 +38,7 @@ import {
   terminalText,
   uninstallHooks,
   writeBaseline
-} from "./chunk-XVV5FNDV.js";
+} from "./chunk-L4XLCXC2.js";
 
 // src/cli.ts
 import { Command } from "commander";
@@ -100,7 +100,7 @@ function isDefaultInvocation(options) {
 var NEXT_STEPS = [
   "",
   pc.bold("Next steps"),
-  `  ${pc.cyan("npx agent-roast hook install")}   stop your coding agent from weakening checks (Claude Code, Codex, Cursor)`,
+  `  ${pc.cyan("npx agent-roast hook install")}   stop your coding agent from weakening checks (Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, Factory Droid)`,
   `  ${pc.cyan("npx agent-roast init --ci")}      add the pull-request check (GitHub Actions)`,
   `  ${pc.cyan("npx agent-roast demo")}           see a scripted example`,
   ""
@@ -505,7 +505,7 @@ program.command("demo").description('See what agent-roast reports for a scripted
   }
 });
 program.command("audit").description("Score the Git history by the shortcuts it adds (the original agent-roast report)").argument("[path]", "Path to git repository (default: current directory)", ".").action((repoPath) => rootAction(repoPath, { ...program.opts(), forceLegacy: true }));
-var hook = program.command("hook").description("Guard a coding agent: send it back to fix the code when it weakens a check (Claude Code, Codex, Cursor)");
+var hook = program.command("hook").description("Guard a coding agent: send it back to fix the code when it weakens a check (Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, Factory Droid)");
 hook.command("run").description("Run by the agent when it is about to finish; reads the agent's JSON on stdin and answers in its dialect").option("--agent <agent>", `${HOOK_AGENTS.join(", ")} (detected from the input when omitted)`).option("--advisory", "only warn the person; never send the agent back", false).action(async (opts) => {
   let agent;
   try {
@@ -525,13 +525,13 @@ hook.command("run").description("Run by the agent when it is about to finish; re
     process.exitCode = 0;
   }
 });
-hook.command("install").description("Add the hook to Claude Code, Codex or Cursor in this repository").argument("[agents...]", `${HOOK_AGENTS.join(", ")} (default: the agents that already have a folder in this repository)`).option("--global", "install for every repository of this user instead", false).option("--local", "Claude Code only: use .claude/settings.local.json, which stays out of Git", false).option("--command <program>", "program to run instead of 'npx --yes agent-roast@<version>', for example 'agent-roast' when installed globally").option("--advisory", "try it without risk: the hook only warns you and never sends the agent back", false).option("--instructions", "also write a short 'keep the checks honest' paragraph into AGENTS.md / CLAUDE.md", false).option("--dry-run", "show what would change without writing anything", false).action((names, opts) => {
+hook.command("install").description("Add the hook to Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot or Factory Droid in this repository").argument("[agents...]", `${HOOK_AGENTS.join(", ")} (default: the agents that already have a folder in this repository)`).option("--global", "install for every repository of this user instead", false).option("--local", "Claude Code only: use .claude/settings.local.json, which stays out of Git", false).option("--command <program>", "program to run instead of 'npx --yes agent-roast@<version>', for example 'agent-roast' when installed globally").option("--advisory", "try it without risk: the hook only warns you and never sends the agent back", false).option("--instructions", "also write a short 'keep the checks honest' paragraph into AGENTS.md / CLAUDE.md", false).option("--dry-run", "show what would change without writing anything", false).action((names, opts) => {
   try {
     if (opts.global && opts.local) throw new ConfigError("Choose --global or --local, not both.");
     let agents = parseAgents(names);
     const root = repositoryRoot(process.cwd());
-    if (!agents.length) agents = HOOK_AGENTS.filter((agent) => fs.existsSync(path.join(root, { claude: ".claude", codex: ".codex", cursor: ".cursor" }[agent])));
-    if (!agents.length) throw new ConfigError("Name the agent to guard: agent-roast hook install claude (or codex, cursor).");
+    if (!agents.length) agents = HOOK_AGENTS.filter((agent) => fs.existsSync(path.join(root, { claude: ".claude", codex: ".codex", cursor: ".cursor", gemini: ".gemini", copilot: ".github/hooks", droid: ".factory" }[agent])));
+    if (!agents.length) throw new ConfigError("Name the agent to guard: agent-roast hook install claude (or codex, cursor, gemini, copilot, droid).");
     const scope = opts.global ? "global" : opts.local ? "local" : "project";
     const advisory = opts.advisory || program.opts().advisory === true;
     const changes = installHooks({ cwd: process.cwd(), agents, scope, command: opts.command, advisory, dryRun: opts.dryRun });
@@ -542,6 +542,9 @@ hook.command("install").description("Add the hook to Claude Code, Codex or Curso
       for (const change of installInstructions({ cwd: process.cwd(), dryRun: opts.dryRun })) console.log(`  + ${displayPath(change.file)} (${change.action}): keep-the-checks-honest paragraph`);
     }
     if (agents.includes("codex")) console.log("\nCodex runs a new hook only after you trust it: open Codex and review it with /hooks.");
+    if (agents.includes("gemini")) console.log("\nGemini CLI warns the first time it sees a project hook, and again when its command changes; /hooks panel shows it.");
+    if (agents.includes("copilot")) console.log("\nGitHub Copilot reads hook files when it starts: restart it. The cloud agent loads them from the default branch only.");
+    if (agents.includes("droid")) console.log("\nFactory Droid takes a snapshot of its hooks at startup: restart it, and review the hook with /hooks.");
     if (!opts.dryRun && advisory) console.log("\nAdvisory mode: when the agent finishes with weakened checks you get a one-line warning and the agent is never sent back.\nStart sending it back: agent-roast hook install (without --advisory)   \xB7   remove it: agent-roast hook uninstall");
     else if (!opts.dryRun) console.log("\nWhen the agent is about to finish with weakened checks it is sent back once to fix the code.\nTurn it off for a while: agent-roast hook pause   \xB7   only warn instead: agent-roast hook install --advisory   \xB7   remove it: agent-roast hook uninstall");
   } catch (error) {
@@ -549,7 +552,7 @@ hook.command("install").description("Add the hook to Claude Code, Codex or Curso
     process.exitCode = 2;
   }
 });
-hook.command("uninstall").description("Remove the hook from Claude Code, Codex or Cursor").argument("[agents...]", `${HOOK_AGENTS.join(", ")} (default: all)`).option("--global", "remove it from the user-level files", false).option("--local", "Claude Code only: remove it from .claude/settings.local.json", false).option("--dry-run", "show what would change without writing anything", false).action((names, opts) => {
+hook.command("uninstall").description("Remove the hook from Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot or Factory Droid").argument("[agents...]", `${HOOK_AGENTS.join(", ")} (default: all)`).option("--global", "remove it from the user-level files", false).option("--local", "Claude Code only: remove it from .claude/settings.local.json", false).option("--dry-run", "show what would change without writing anything", false).action((names, opts) => {
   try {
     if (opts.global && opts.local) throw new ConfigError("Choose --global or --local, not both.");
     const agents = names.length ? parseAgents(names) : [...HOOK_AGENTS].filter((agent) => !(opts.local && agent !== "claude"));
@@ -565,11 +568,11 @@ hook.command("uninstall").description("Remove the hook from Claude Code, Codex o
 hook.command("status").description("Show where the hook is installed").action(async () => {
   try {
     const rows = hookStatus({ cwd: process.cwd() });
-    for (const row of rows) console.log(`  ${row.installed ? pc.green("installed") : pc.dim("not set  ")}  ${AGENT_LABELS[row.agent].padEnd(11)} ${row.scope.padEnd(8)} ${displayPath(row.file)}${row.advisory ? pc.yellow("  (advisory: warns, never sends the agent back)") : ""}`);
+    for (const row of rows) console.log(`  ${row.installed ? pc.green("installed") : pc.dim("not set  ")}  ${AGENT_LABELS[row.agent].padEnd(16)} ${row.scope.padEnd(8)} ${displayPath(row.file)}${row.advisory ? pc.yellow("  (advisory: warns, never sends the agent back)") : ""}`);
     const paused = await hookPausedUntil(process.cwd());
     if (paused) console.log(`
   Paused in this repository until ${paused.toLocaleString()} (agent-roast hook resume).`);
-    if (!rows.some((row) => row.installed)) console.log("\n  Install it with: agent-roast hook install claude   (or codex, cursor)");
+    if (!rows.some((row) => row.installed)) console.log("\n  Install it with: agent-roast hook install claude   (or codex, cursor, gemini, copilot, droid)");
   } catch (error) {
     console.error(`agent-roast hook status error: ${errorMessage(error)}`);
     process.exitCode = 2;

package/dist/index.d.ts

@@ -475,7 +475,7 @@ declare function replayHistory(options?: ReplayOptions): Promise<ReplayResult>;
 declare function formatReplayTerminal(result: ReplayResult, allRules?: boolean): string;
 
 /** Coding agents that can run a command when they are about to finish and feed its answer back. */
-type HookAgent = "claude" | "codex" | "cursor";
+type HookAgent = "claude" | "codex" | "cursor" | "gemini" | "copilot" | "droid";
 declare const HOOK_AGENTS: readonly HookAgent[];
 interface HookOptions {
     /** Which agent is calling. Detected from its input when omitted. */

package/dist/index.js

@@ -92,7 +92,7 @@ import {
   unquoteGitPath,
   validateExceptions,
   writeBaseline
-} from "./chunk-XVV5FNDV.js";
+} from "./chunk-L4XLCXC2.js";
 export {
   AGENT_LABELS,
   CONFIG_FILENAMES,

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "agent-roast",
-  "version": "0.21.0",
-  "description": "Stops AI coding agents (and rushed humans) from turning a build green by weakening tests, coverage gates or CI. Hook for Claude Code, Codex and Cursor, pre-commit, GitHub Action. Offline.",
+  "version": "0.23.0",
+  "description": "Stops AI agents (and rushed humans) weakening tests, coverage gates or CI to get a green build. Hook for Claude Code, Codex, Cursor, Gemini CLI, Copilot, Droid; pre-commit; GitHub Action. Offline.",
   "mcpName": "io.github.Nari-Ab/agent-roast",
   "type": "module",
   "main": "./dist/index.js",
@@ -34,6 +34,8 @@
     "codex",
     "cursor",
     "copilot",
+    "gemini-cli",
+    "factory-droid",
     "hooks",
     "guardrails",
     "code-review",