devlyn-cli 4.3.30 → 4.3.32

package/AGENTS.md

@@ -13,7 +13,7 @@ Seven rules govern every change. Cite them by name when a decision touches one.
 1. **No workaround** — fix the root cause, never the symptom. No `any`, no `@ts-ignore`, no silent `catch`, no hardcoded fallback that hides a broken contract. No config bypasses.
 2. **No overengineering** — smallest sufficient change, one responsibility per module. New abstractions need an observed failure. Subtractive-first: ask "what can I delete instead?" before writing anything new. Remove code your change makes unused; only report unrelated pre-existing dead code.
 3. **No guesswork** — verify with the actual files, logs, diffs, and run output before forming conclusions. State the falsifiable prediction BEFORE the experiment; record raw results AFTER. Retroactive prediction edits are dishonest.
-4. **Worldclass** — zero CRITICAL, zero HIGH security/design findings on the shippable path. Any review is one pass in which every reviewer, including each one the user requires, reads the source the checks ran on; fix each finding or reject it with a reason, re-check the fixes and the invariants they touch, and repeat the full review only while a CRITICAL or HIGH finding remains.
+4. **Worldclass** — zero CRITICAL, zero HIGH findings on the shippable path; each must name a concrete case of wrong behavior or results, a security hole, or lost or corrupted data. A review is one pass on the source the checks ran on, by the reviewers the user requires, else by one reviewer who has not seen the author's reasoning. Fix each finding or reject it with a reason, and verify every edit and the invariants it touches. Only a CRITICAL or HIGH calls for another reviewer pass, which reads the fixes, the rejections and the invariants they touch. If one survives that pass, or its fix caused a new one, repair once more; if the next pass finds either again, stop: the work does not ship and is reported blocked.
 5. **Best practice** — idiomatic for the language and framework. Use standard primitives; do not hand-roll what the library already provides.
 6. **Optimized** — efficient on the resource that matters (the user's time, tokens, attention, runtime) without weakening correctness or requirements. An extra check or process step must beat the simpler path.
 7. **Production ready** — error states are explicit and visible; behavior under failure is what the user expects, not silent corruption.

package/CLAUDE.md

@@ -13,7 +13,7 @@ Seven rules govern every change. Cite them by name when a decision touches one.
 1. **No workaround** — fix the root cause, never the symptom. No `any`, no `@ts-ignore`, no silent `catch`, no hardcoded fallback that hides a broken contract. No config bypasses.
 2. **No overengineering** — smallest sufficient change, one responsibility per module. New abstractions need an observed failure. Subtractive-first: ask "what can I delete instead?" before writing anything new. Remove code your change makes unused; only report unrelated pre-existing dead code.
 3. **No guesswork** — verify with the actual files, logs, diffs, and run output before forming conclusions. State the falsifiable prediction BEFORE the experiment; record raw results AFTER. Retroactive prediction edits are dishonest.
-4. **Worldclass** — zero CRITICAL, zero HIGH security/design findings on the shippable path. Any review is one pass in which every reviewer, including each one the user requires, reads the source the checks ran on; fix each finding or reject it with a reason, re-check the fixes and the invariants they touch, and repeat the full review only while a CRITICAL or HIGH finding remains.
+4. **Worldclass** — zero CRITICAL, zero HIGH findings on the shippable path; each must name a concrete case of wrong behavior or results, a security hole, or lost or corrupted data. A review is one pass on the source the checks ran on, by the reviewers the user requires, else by one reviewer who has not seen the author's reasoning. Fix each finding or reject it with a reason, and verify every edit and the invariants it touches. Only a CRITICAL or HIGH calls for another reviewer pass, which reads the fixes, the rejections and the invariants they touch. If one survives that pass, or its fix caused a new one, repair once more; if the next pass finds either again, stop: the work does not ship and is reported blocked.
 5. **Best practice** — idiomatic for the language and framework. Use standard primitives; do not hand-roll what the library already provides.
 6. **Optimized** — efficient on the resource that matters (the user's time, tokens, attention, runtime) without weakening correctness or requirements. An extra check or process step must beat the simpler path.
 7. **Production ready** — error states are explicit and visible; behavior under failure is what the user expects, not silent corruption.

package/README.md

@@ -72,7 +72,9 @@ A bare intent is planned and, when your request authorizes the work, added and d
 
 Direct work edits your current checkout; run parallel sessions in separate worktrees (for example `claude -w`). Completed, verified direct requests with changes, and
 every drained task, are delivered by project policy from their own linked worktrees. A commit request stays local and a PR request stops at the PR; otherwise accepted tasks default to
-scoped commit → push → PR → merge when the repository allows it (otherwise the PR waits for a person). After the
+scoped commit → push → PR → merge when the repository allows it (otherwise the PR waits for a person; while required
+checks or reviews have not passed it stays pending, and running the printed resume after they pass merges it, or a person
+merges it; with GitHub's Allow auto-merge on, delivery requests the merge, so it lands by itself). After the
 PR merges, the worktree and branch its session released are cleaned; anything in use is
 kept. Set `git config --local devlyn.completionMode pr` to stop at the PR; `task-complete.py complete --mode auto|pr`
 overrides one task. Local-only/no-push instructions take precedence: nothing is pushed or merged;

package/bin/devlyn.js

@@ -616,23 +616,46 @@ function installLocalSkill(skillName, roots, installedSkills = new Map()) {
   return true;
 }
 
-// One skill from `source` into one skill-loader directory. The copy is staged and checked first, so
-// a failed copy leaves any older copy of it in place for a plain rerun. A copy under its name
+// One skill from `source` into one skill-loader directory. The copy is staged and checked first, and
+// the older copy is moved aside into the stage before the new one is renamed in, and back if that
+// fails, so a failed copy or swap leaves the older copy in place for a plain rerun. If moving it back
+// or removing it after the swap fails, the error names where it is kept. A copy under its name
 // before the 4.0.0 rename goes only once the new one is complete.
 function installSkillInto(source, target, skillName, installedSkills) {
   const dest = path.join(target, skillName);
   const stage = fs.mkdtempSync(path.join(target, '.devlyn-skill-'));
   const stagedSkill = path.join(stage, skillName);
+  const previous = path.join(stage, 'previous');
+  let files;
+  let moved = false;
   try {
-    const files = copyRecursive(path.join(source, skillName), stagedSkill, stage);
+    files = copyRecursive(path.join(source, skillName), stagedSkill, stage);
     assertCompleteSkillInstall(source, stage, [skillName]);
-    fs.rmSync(dest, { recursive: true, force: true });
+    if (fs.lstatSync(dest, { throwIfNoEntry: false })) {
+      fs.renameSync(dest, previous);
+      moved = true;
+    }
     fs.renameSync(stagedSkill, dest);
-    installedSkills.set(dest, files.map((file) => path.join(dest, path.relative(stagedSkill, file))));
-  } finally {
+  } catch (error) {
+    if (moved) {
+      try { fs.renameSync(previous, dest); }
+      catch (rollback) {
+        throw new AggregateError([error, rollback], `${error.message}; ${rollback.message}; the previous copy is kept at ${previous}`);
+      }
+    }
+    try { fs.rmSync(stage, { recursive: true, force: true }); }
+    catch (cleanup) {
+      throw new AggregateError([error, cleanup], `${error.message}; ${dest} is as it was, but ${stage} could not be removed: ${cleanup.message}`);
+    }
+    throw new Error(`${error.message}; ${dest} is as it was`, { cause: error });
+  }
+  installedSkills.set(dest, files.map((file) => path.join(dest, path.relative(stagedSkill, file))));
+  try {
     fs.rmSync(stage, { recursive: true, force: true });
+  } catch (error) {
+    const kept = fs.lstatSync(previous, { throwIfNoEntry: false }) ? `its previous copy at ${previous}` : stage;
+    throw new Error(`${dest} is installed, but ${kept} could not be removed: ${error.message}`);
   }
-
 }
 
 // A 0.6.0-0.7.1 copy is one SKILL.md, the file that identifies it. A failed swap leaves
@@ -2256,7 +2279,9 @@ function cleanupDeliver(scope, items, current) {
   if (!prepared.acceptance) return cleanupUpdateCheckout(scope);
   const completed = lastJson(run(['complete', '--receipt', allocated.receipt, '--acceptance', prepared.acceptance, '--writers-stopped'], scope.root).stdout);
   const status = completed?.delivery_status || completed?.status;
-  if (status === 'COMPLETE' || status === 'MERGED') return { ...cleanupUpdateCheckout(scope), delivery: { state: 'merged', url: completed.pr || null } };
+  // A merge whose workspace or scratch the helper retained (CLEANUP_PENDING) is not a clean merge: report its resume.
+  if (status === 'COMPLETE' || status === 'MERGED') return { ...cleanupUpdateCheckout(scope), delivery: completed.status === 'CLEANUP_PENDING'
+    ? { state: 'cleanup-pending', url: completed.pr || null, resume: completed.resume } : { state: 'merged', url: completed.pr || null } };
   if (completed?.pr) return { delivery: { state: 'pr-open', url: completed.pr } };
   const reason = completed?.reason || 'delivery did not complete';
   return { error: cleanupReason(failureCode(reason), { detail: reason, ...(completed?.resume ? { resume: completed.resume } : {}) }) };

package/bin/harness-activation.js

@@ -380,7 +380,10 @@ function inspectSkills(context, release, env, config, active) {
     for (const name of SKILLS) {
       const dir = path.join(root, name);
       try {
-        if (!fs.lstatSync(dir, { throwIfNoEntry: false })) continue;
+        // A lookup through a file finds no entry: POSIX says ENOTDIR where Windows says "not found".
+        let entry;
+        try { entry = fs.lstatSync(dir, { throwIfNoEntry: false }); } catch (error) { if (error.code !== 'ENOTDIR') throw error; }
+        if (!entry) continue;
         const hash = fingerprint(dir);
         if (!hashes.has(name)) hashes.set(name, fingerprint(path.join(release.root, 'config/skills', name)));
         found.push({ name, path: path.join(dir, 'SKILL.md'), owned: hash === hashes.get(name) || (history.skills[name] || []).includes(hash) });

package/bin/harness-status.js

@@ -70,16 +70,23 @@ function requirements() {
   const probes = [
     ['python', true, 'python3', ['-B', '--version']],
     ['git', true, 'git', ['--version']],
-    ['git-bash', process.platform === 'win32', 'bash', ['--version']],
-    ['gh', false, 'gh', ['--version']],
+    // Git Bash is a bash whose builtin `pwd -W` (an MSYS2 patch) prints a Windows path; its version label no longer
+    // tells it from Cygwin's. -p reads no BASH_ENV or ENV file; --norc no ~/.bashrc, which a bash at shell level 1
+    // otherwise reads when SSH_CLIENT is set or stdin is a socket (Node's pipes are). Long options come first. macOS
+    // and Linux bashes refuse -W, so there this row (never required) is unavailable.
+    ['git-bash', process.platform === 'win32', 'bash', ['--norc', '-p', '-c', 'echo "$BASH_VERSION"; cd / && pwd -W']],
+    // gh 2.91 and later writes a telemetry device id into its state folder on every command unless this is off.
+    ['gh', false, 'gh', ['--version'], { GH_TELEMETRY: 'false' }],
   ];
-  return [node, ...probes.map(([name, required, binary, argv]) => {
-    const result = spawnSync(binary, argv, { encoding: 'utf8', timeout: 5000, env: { ...process.env, PYTHONDONTWRITEBYTECODE: '1' } });
+  return [node, ...probes.map(([name, required, binary, argv, env]) => {
+    const result = spawnSync(binary, argv, { encoding: 'utf8', timeout: 5000, env: { ...process.env, PYTHONDONTWRITEBYTECODE: '1', ...env } });
     const output = (result.stdout || result.stderr || '').trim();
-    const match = output.match(/\b(\d+)\.(\d+)(?:\.\d+)?/);
-    let available = !result.error && result.status === 0 && Boolean(match);
+    const bash = name === 'git-bash';
+    // A bash's version is its $BASH_VERSION line, never digits in the path line after it.
+    const match = output.match(bash ? /^\d+\.\d+\.\d+(?=(?:\(\d+\)-\w+)?\r?$)/m : /\b(\d+)\.(\d+)(?:\.\d+)?/);
+    let available = !result.error && result.status === 0
+      && (bash ? /^(?:[A-Za-z]:\/|\/\/)/.test(output.split('\n').pop()) : Boolean(match));
     if (name === 'python' && available) available = Number(match[1]) > 3 || (Number(match[1]) === 3 && Number(match[2]) >= 11);
-    if (name === 'git-bash' && required && available) available = /mingw|msys/i.test(output);
     return { name, required, available, version: match?.[0] || null,
       reasons: available ? [] : [{ code: 'requirement-unavailable', params: { name } }] };
   })];
@@ -168,4 +175,4 @@ function statusCommand(args) {
   return status;
 }
 
-module.exports = { harnessStatus, statusCommand, repositoryIdentity, inspectInstruction };
+module.exports = { harnessStatus, statusCommand, repositoryIdentity, inspectInstruction, requirements };

package/bin/instruction-templates.json

@@ -469,6 +469,7 @@
       "bddb4f88a0c11165b9cafe7a31027e9769313e1aee7c0a9ed9dad490e4fe1bcc",
       "be11c198a50be2884c4c1efec52bbddff55c7547a5daf0ec2ef046279c95dc45",
       "bfdb7c7e4db21e1c452e730cecadc8e97abd317c6e54ae58c298f4288b1f1354",
+      "c1970e6a96f0c4241907d29fb4cdf831ad255b2bbfdb5270fe0aefc32cb91f0d",
       "c1a1e9722a0f3c13c693eeaacdac8cc3d13d36d242f2bab58311170496ea4e2e",
       "c236da36a45d4013ffb7e131575229b605f662a30b5174f14754f519acc24252",
       "c276962617888c5ac2d352c5e2ed018c5ce9d00f199c73e1dd61438ec4a914fb",
@@ -997,6 +998,7 @@
       "c115b2ba9ec7abe2fa5261a5894e3c4899794b527f01ec682ae7fd2479199c08",
       "c150c40a04099d7b726eb8aecc8eec8e92f9f9248a6a5c15556c08000c4bd231",
       "c16ff4fe013ee2cef905d767a09305d9c1b2f778a06b439f0bc4645ac05bc82b",
+      "c1970e6a96f0c4241907d29fb4cdf831ad255b2bbfdb5270fe0aefc32cb91f0d",
       "c1c7e45fedafb97770b6726539694eb0668c70677c2031fe2db2bc8b5339423a",
       "c227759474dc46ec6df00fe7baf8ba4b6b0adc71f5068ef522ff46526c861a48",
       "c27b13c3ae4e6fb34d9b28b8f4e899e29bfbf606a69207b9532d0e44e6078a2e",

package/bin/skill-history.json

@@ -12,6 +12,7 @@
       "08c91060cfec04518de582599d1892ad50f6e6cec69aecd8678760c3e0039fa0",
       "0a60dd9b96442be24a571dd5fe6cac426bba7dd7a06ad17f3211f4f9fd9b1741",
       "0b5c3f72d18f8eea538bfef7192b207ff5cbac25c7e0d2ce080cc725bfdaa2e2",
+      "0d7e8097f36623f68b7cdc26e9c2eb80d9792b3776edf308d306febdfd9f26ae",
       "0f397a4edd6b1b3063d27f6205dc6b3d069fa616624bccf661539b174535fc7e",
       "110005e3c7052d35502186d907b169c9fc7c85d558ab7c7ad0ba2f0308aa7a5d",
       "11ac38b21ecc139211af1b335bb3acc4f0f1bc4b5c133e81a80cafebcaaca298",
@@ -56,6 +57,7 @@
       "406b982f6ca0b9f5c0faa6cb50749b52aaf23545e06174ec39f831929dc929bd",
       "40f50f3c3c0d030d1efeb6955c0c6f32f518e737d8dd3bec6610ff02badebbdb",
       "430d156eb8c62c7edba3d660050f90b267f928976c3a3688e95d54f00dc34747",
+      "446e47fc34c81ed4cb9f4332d446a93025ae3dc1791155a4b80d6f195255958e",
       "450d578ef177a9e8f0b693940ad64b41c1d9719975ba6305f62a7c82ac3a892f",
       "45ef27af9ec27089e2861c13b882521fc6bf073dc6b7a356b326b4a27c5e6e24",
       "4685be6a94ad5587ce53785c14a60cccf9a852d9d23391dba0a5bcd0c64931cf",
@@ -119,6 +121,7 @@
       "8c80926ab03b2477e52264495b35c2116f65fc51b02e8c56e441b5705872392a",
       "8fb73c0d2472b3643ac6e3f0ad87d818694afebaf68fcad596496e703893c43d",
       "9162fc2fed6f4a611d299485be3a397080beb3304865ce59b95576ba7d00650e",
+      "92ffe78e156d8f11f666a92b26d21ae52e147ed9cad7eb79efbef334a9373281",
       "94a14a1b87df1d7297d12949a24e47a0d6c16a8fe30495ee759fcabb2e1aa398",
       "978e4210e49b42827c98cf6e40c868c87dc6e3533a2b779ca2fa05abfc16ad0f",
       "97e7ad89c74c609b15dfbf6f9f7ac61df8b50e97e4bc318f72922e058ecffba4",
@@ -196,6 +199,7 @@
       "f2bcfa0471aba81a10e3d0e6fd43bd6ec670601a4ba3a7d719e595edf9ab032a",
       "f958bfabd2e056fea911526fba2196d2ed5a48828f576e5ce1657100b466db20",
       "f9617e1034543f31ae2a3820301fbac6e4674e5d6fbc0d4b76751fe87b8e013e",
+      "fbca65e47854425acd80087e89024fb9d2f44f853d88abe582fbee107a3e8de3",
       "fd68c7a74272929a14e685e447a51058ffcdd4c1d6bef83cbba18045c07aee55",
       "fdb608b34a8600b4ce209d7f645ec5ad4475948be6ad95c941cc7ef02b5959df",
       "fe357defe8ac6efd6bbd15b3bf35f79d79761c0f56d507488d9428536d4fb5f7",
@@ -224,7 +228,8 @@
       "3fe0742825aff05fa6fda49d3e3e0fbf5457307af944eb90f2771cbf611edd69",
       "4a8bed603191eeb6834a5008691bb3942e77462abd2818340b808a752ee5a224",
       "57c2481f9ae36623f59b365eae82f22d54253f8ba0fd690949a12858b5f6e308",
-      "b352d21d2a86def2997d89ae1ffa35b0fe428781f75cfa9e1b69a3384416ec31"
+      "b352d21d2a86def2997d89ae1ffa35b0fe428781f75cfa9e1b69a3384416ec31",
+      "be2562b5541d754c34e42b4242751bbaf3b3b2780f4314552ebb63a350e23170"
     ],
     "devlyn-auto-resolve": [
       "0740bb91fb99953631b0c2c273f00210544576f9497af5eecfdd5fe1e58349e2",
@@ -363,6 +368,8 @@
       "36b58dbdc26e51abc05a60872c4582faeebd67e9fe5626a3933467ba07c68dc4",
       "40237cd5259dbcf7c490f130e844a9e32eea48a8edabd50800edee5029bf95a4",
       "4213143689f7e7fe62ab8743c256a45492fafffc1757d64901b7b02725aee5c3",
+      "437659470cb978840f503c7d4a2af75c78beb2752cc3eed413bb33de346bd561",
+      "472d2ad853c96651324aa9a3de00cb61c0cecacb389e850ab1b380d535d22922",
       "49d6e024b343dd77e4c0cea0e513e40fe679a7a8192953190d042b3a21bbf69c",
       "5536f8391ec4442f1b793dba8dd5d7820d45f4b6dadf1111cbac9d54469a01ca",
       "573a6059a7383b37affd9c2e8bc6c88a4fb54117628a5a0746406cc32111e9aa",
@@ -381,6 +388,7 @@
       "6e2b54571bfbc1fab571b81564fb0625ed83333fc12277c473fdf5a231f7d40c",
       "6e7228610628fc4bca31dc0b0aaf5eb4445e1f123f2c54bd13921c5cd01129fe",
       "75cecbce4d1f4adbba608cef498ae7e421ea75512c6bf9282b66cd94d821f672",
+      "78be1509ca8cf797f8c6e790d0d49ab608e50e0b828916c8c6ee790165ce792b",
       "7f451e48e391643b34f6249244bc47d93d6303875e7b092771e3af0718343bc1",
       "7fb3ceaf835230e1a75e510b046f4fe26e79815760354bec1949fa55515c6316",
       "83e9de10d3c439af17dc4d7a603109d35136c5c8bd468ab2be962e1fad064c3c",
@@ -404,6 +412,7 @@
       "db8d921aabf0d04811b55465a3757c62c4b39631aebaf6eecc994ca241cfbe56",
       "e25da746dc37b2e82c658a28d9481fd69741151045683eee2119582ddaf731cd",
       "e47a2ce766ae81fa0870e0a8dfbfe737311de69ce620e92217abbe4c532e3c8b",
+      "e67d3662836adf13c2e7177553bba6e7695484bd40438e8188cb9dd6d956cd2f",
       "e929f76c15349f1b9bc1e84e0a3e9e48e916168d79270e0f4a36ea8eab9b5388",
       "ebc693658da0b260c876f29cd6dfc2b12ae6690e3532622fa9c1cd4481162bd8",
       "f2243cc4150e225961825f6097b0cf38166c291dcf58127516ceee841f7ed4c8",
@@ -825,6 +834,8 @@
     "4.3.29",
     "4.3.3",
     "4.3.30",
+    "4.3.31",
+    "4.3.32",
     "4.3.4",
     "4.3.5",
     "4.3.6",

package/config/skills/_shared/bring-in.py

@@ -363,6 +363,8 @@ class BringInTests(unittest.TestCase):
         self.env = {key: value for key, value in os.environ.items()
                     if not key.startswith("GIT_")}
         self.env.update(GIT_CONFIG_NOSYSTEM="1", GIT_CONFIG_GLOBAL=str(self.root / "no-global"),
+                        # A commit's detached automatic maintenance would still write into .git while a test copies it.
+                        GIT_CONFIG_COUNT="1", GIT_CONFIG_KEY_0="maintenance.auto", GIT_CONFIG_VALUE_0="false",
                         GIT_AUTHOR_NAME="Fixture", GIT_AUTHOR_EMAIL="fixture@example.invalid",
                         GIT_COMMITTER_NAME="Fixture", GIT_COMMITTER_EMAIL="fixture@example.invalid",
                         LC_ALL="C", GIT_EDITOR=":", GIT_MERGE_AUTOEDIT="no")
@@ -415,6 +417,14 @@ class BringInTests(unittest.TestCase):
         path.write_bytes(("#!/bin/sh\n" + body).encode("utf-8"))
         path.chmod(0o755)
 
+    def test_fixture_commit_starts_no_automatic_maintenance(self):
+        # Its detached child would still write .git/objects/maintenance.lock while copytree or cleanup reads the tree.
+        trace = self.root / "trace.json"
+        self.env["GIT_TRACE2_EVENT"] = str(trace)
+        self.commit("traced", "traced\n")
+        children = [line for line in trace.read_text(encoding="utf-8").splitlines() if '"event":"child_start"' in line]
+        self.assertEqual(children, [])
+
     def test_fast_forward_and_paths_with_spaces(self):
         self.cli()
         self.assertEqual(self.g("rev-parse", "HEAD").strip(), self.target)

package/config/skills/_shared/runtime-principles.md

@@ -15,7 +15,7 @@ Seven rules govern every change. Cite them by name when a decision touches one.
 1. **No workaround** — fix the root cause, never the symptom. No `any`, no `@ts-ignore`, no silent `catch`, no hardcoded fallback that hides a broken contract. No config bypasses.
 2. **No overengineering** — smallest sufficient change, one responsibility per module. New abstractions need an observed failure. Subtractive-first: ask "what can I delete instead?" before writing anything new. Remove code your change makes unused; only report unrelated pre-existing dead code.
 3. **No guesswork** — verify with the actual files, logs, diffs, and run output before forming conclusions. State the falsifiable prediction BEFORE the experiment; record raw results AFTER. Retroactive prediction edits are dishonest.
-4. **Worldclass** — zero CRITICAL, zero HIGH security/design findings on the shippable path. Any review is one pass in which every reviewer, including each one the user requires, reads the source the checks ran on; fix each finding or reject it with a reason, re-check the fixes and the invariants they touch, and repeat the full review only while a CRITICAL or HIGH finding remains.
+4. **Worldclass** — zero CRITICAL, zero HIGH findings on the shippable path; each must name a concrete case of wrong behavior or results, a security hole, or lost or corrupted data. A review is one pass on the source the checks ran on, by the reviewers the user requires, else by one reviewer who has not seen the author's reasoning. Fix each finding or reject it with a reason, and verify every edit and the invariants it touches. Only a CRITICAL or HIGH calls for another reviewer pass, which reads the fixes, the rejections and the invariants they touch. If one survives that pass, or its fix caused a new one, repair once more; if the next pass finds either again, stop: the work does not ship and is reported blocked.
 5. **Best practice** — idiomatic for the language and framework. Use standard primitives; do not hand-roll what the library already provides.
 6. **Optimized** — efficient on the resource that matters (the user's time, tokens, attention, runtime) without weakening correctness or requirements. An extra check or process step must beat the simpler path.
 7. **Production ready** — error states are explicit and visible; behavior under failure is what the user expects, not silent corruption.