gm-skill 2.0.1632 → 2.0.1634

package/gm-plugkit/instructions/browser.md

@@ -26,6 +26,8 @@ capture\n<expression>
 profile\n<expression>
 profile interval=<us> topN=<n>\n<expression>
 trace\n<expression>
+screenshot\n<expression>
+dom=<css-selector>\n
 ```
 
 **Open on the page you want to test, not a blank one.** A bare `https://...` URL body navigates the session straight to that page and returns `{url, title}` -- the simplest "show me this page." `url=<url>\n<expression>` navigates first, then runs your expression on the loaded page, so the global/DOM you assert is already there in one dispatch instead of a blank surface you must `page.goto` yourself. `url=` composes with `timeout=` and `capture` -- stack the prefix lines in order `timeout=`, then `url=`, then `capture`, the expression last; the prepended `page.goto` rides inside the capture so its navigation console/network is captured too. A bare expression with no `url=`/bare-URL prefix runs against whatever the session is already on -- a never-navigated session is on `about:blank`, so the expression evaluates an empty page and the envelope comes back with `landed_on_blank: true` and a `hint` telling you to add `url=`; navigate first and the surprise never happens. `session new` returns the id you carry. (`session close` and `session kill` are aliases.) Default per-eval timeout 120000ms; operations that legitimately exceed it prefix `timeout=<ms>\n` (wrapper clamps to 120000ms). The response carries `timeout_ms_used`; `browser.runner-timeout` fires at the cap -- read `stderr`, narrow or raise, never retry blind at the same budget.
@@ -36,9 +38,17 @@ trace\n<expression>
 
 **`trace\n<expression>` catches GPU activity the CPU sampler is structurally blind to.** A V8 CPU profiler samples on-CPU JS call stacks only; GPU-process work -- compositor, raster, draw, WebGL/canvas -- never appears in it. `trace` opens a CDP `Tracing` session over wrapper-controlled categories (`gpu`, `viz`, `cc`, `blink`, `devtools.timeline`), runs your script, ends tracing, and returns `{result, trace: {wall_us, gpu_us, viz_us, cc_us, raster_us, event_count, complete, by_category}, trace_error, debug: {...}}`. `gpu_us`/`viz_us`/`cc_us` are wall-clock microseconds of GPU-process activity summed from the trace; `by_category` is the bounded top-15 category rollup (raw events never returned). When wall greatly exceeds CPU self-time, `trace` is how you attribute the gap to the GPU rather than guessing. `debug.performance` carries paint/frame metrics (`first_contentful_paint_ms`, `largest_contentful_paint_ms`, `cumulative_layout_shift`, `longtasks`, `fps`) for client-side render jank. `tracingComplete` is bounded by a timeout; on a CDP failure `trace_error` is set and `result`/`debug` still return.
 
+**Every dispatch returns a structured `result` -- read `result.foo`, never regex the stdout.** Whatever your expression returns is JSON-serialized into the envelope's top-level `result` field (the human-readable `[return value]` text stays in `stdout` for eyeballing). `capture`/`profile`/`trace` put their `{result, debug, ...}` object there whole, so `result.debug.performance.heap_used_mb` and `result.trace.gpu_us` are direct reads. `debug.performance` now also carries `heap_used_mb`/`heap_total_mb` (JS heap) for leak-hunting; `debug.network` entries carry `method` and `status` and are sorted slowest-first before the 30-cap so the worst request survives; `debug.console` is capped at 50 with a trailing dropped-count note so a chatty page cannot flood the envelope.
+
+**`screenshot\n<expression>` writes a PNG and returns its path.** Runs your expression, then `page.screenshot` to `.gm/witness/` (basename-sanitized, confined to that dir), returning `{result, screenshot_path, screenshot_error}`. Reach for it when the bug is visual and the DOM/console do not show it.
+
+**`dom=<css-selector>\n` is the zero-boilerplate element probe.** Returns `{selector, match_count, elements:[{tag, text, attrs, visible, rect}]}` for up to 20 matches -- the fastest answer to "is this element there and what does it say." An invalid selector returns `result.error` (no crash). Composes with `url=`.
+
+**One session per run -- reuse it, then close it.** A browser session is keyed by its spool `sessionId`; every dispatch carrying the SAME sessionId reuses the SAME chromium. A DIFFERENT sessionId opens its OWN chromium -- so a run that invents `probe`/`w2`/`w3`/... names leaks one browser per name. Pick one sessionId, use it for every dispatch, and end with `session close` so nothing is left open; the eval envelope carries a `multi_session_warning` the moment a second distinct session opens. The idle reaper (closes sessions unused past the idle window) and the OS-orphan reaper (kills managed chromiums no live session owns, sparing in-use ones and your own Chrome) are backstops for crashes, not a license to leave sessions open -- close yours.
+
 ## Envelope
 
-`{ok, stdout, stderr, exit_code, session_id?, navigation_requested, landed_on_blank?, hint?}`. `stdout` = stringified eval result; `stderr` = page errors + launch diagnostics; `exit_code` non-zero = the dispatch did not land -- read `stderr` and re-dispatch, never blind. `navigation_requested` reflects whether the dispatch carried a `url=`/bare-URL navigation; `landed_on_blank: true` with a `hint` means the expression ran against `about:blank` -- prefix `url=<target>` and re-dispatch.
+`{ok, stdout, stderr, exit_code, session_id?, navigation_requested, landed_on_blank?, hint?, multi_session_warning?}`. `stdout` = stringified eval result; `stderr` = page errors + launch diagnostics; `exit_code` non-zero = the dispatch did not land -- read `stderr` and re-dispatch, never blind. `navigation_requested` reflects whether the dispatch carried a `url=`/bare-URL navigation; `landed_on_blank: true` with a `hint` means the expression ran against `about:blank` -- prefix `url=<target>` and re-dispatch.
 
 ## Headed by default
 

package/gm-plugkit/instructions/execute.md

@@ -38,7 +38,7 @@ First emit = closure of the transform; scaffold + IOU externalizes residual cost
 
 Data first -- get the structures and their invariants right and the code writes itself; convoluted control flow means the data model is wrong, so fix the model. Make invalid state unrepresentable -- pass parameters over hidden globals, encode the constraint in the type/shape so the bad combination cannot be constructed. Reason from physical constraints (latency, bandwidth, memory, coordination, the worst node) before designing within them. Keep the spine flat, each unit single-focus and understandable at its call site. Make misuse structurally impossible, not documented-against. Optimize the worst case, not the average; design every failure path explicitly (full -> degraded -> safe-fail -> explicit-error), never a silent catastrophic mode. Measure, do not assume -- profile before optimizing, implement both and compare on real input when in genuine dispute. When a change regresses something that worked, revert first and investigate second: restore green, then diagnose from a known-good base. Fail fast and loud over limping on bad state.
 
-**Process of elimination is the debugging paradigm on every surface, and manual labour against real services is how you witness.** This is thinking-in-code at its sharpest: each candidate cause is a hypothesis, and you test the hypothesis by running it, not by reasoning around it. Never guess-and-restart, a/b-test, or shotgun variants: enumerate the candidate causes as mutables, then eliminate each by a witness read against REAL input -- `exec_js` against the real service, `codesearch`/`Read` against the real source, the `browser` verb's `page.evaluate` against a `window.*` global on the live page. Each elimination reveals the next mutable; record it and keep going until one cause survives every other's refutation. Reading the live runtime once observes more than a hundred blind restarts. Profile on the real surface, not from intuition: wrap the suspect node and read the live numbers. In node, `exec_js` carries `duration_ms` for free, surfaces your own timing and `process.memoryUsage()` on stdout, and lands the thrown-error `stack` on stderr -- read both channels (numbers on stdout, stack on stderr). In the browser, a body prefixed `capture\n<script>` auto-returns `{result, debug:{console, pageErrors, network, performance}}` with zero boilerplate. When the slow node is not obvious, sample it bottom-up: `exec_js` with `opts.profile:true` and the browser `profile\n<script>` prefix both return `{result, profile:{timeframe:{start_us,end_us,total_us,sample_count}, culprits:[{location,function,self_us,self_pct,hits}]}}` -- the worst-N `file:line` by self-time across init and code-execution, identical shape on both surfaces, so the culprit ranking points straight at the line to fix. Both also return `mem` (rss/heap/delta) and `wall_vs_cpu:{wall_us, offcpu_us}` -- the sampler sees only on-CPU JS, so a large `offcpu_us` means the time is going to IO, async wait, or the GPU, not the JS you can see; tune with `opts.sampleIntervalUs`/`opts.profileTopN` (cli) or `interval=`/`topN=` (browser). The CPU sampler is structurally blind to GPU activity -- when wall greatly exceeds CPU self-time on a render/canvas/WebGL surface, the browser `trace\n<script>` prefix opens CDP Tracing and returns `trace:{wall_us, gpu_us, viz_us, cc_us, by_category}`, the wall-clock GPU-process time the profiler cannot show. Profile to LOCATE the slow/broken node, then eliminate hypotheses by live measurement. Verification is the same labour: run the real thing and witness the real output (the single mock-free `test.js`, the live page, the real service), never an automated unit/mock harness standing in for the real-services witness. Apparent tooling failure is part of this -- it is your mechanical self-recovery by elimination, never a question for the user.
+**Process of elimination is the debugging paradigm on every surface, and manual labour against real services is how you witness.** This is thinking-in-code at its sharpest: each candidate cause is a hypothesis, and you test the hypothesis by running it, not by reasoning around it. Never guess-and-restart, a/b-test, or shotgun variants: enumerate the candidate causes as mutables, then eliminate each by a witness read against REAL input -- `exec_js` against the real service, `codesearch`/`Read` against the real source, the `browser` verb's `page.evaluate` against a `window.*` global on the live page. Each elimination reveals the next mutable; record it and keep going until one cause survives every other's refutation. Reading the live runtime once observes more than a hundred blind restarts. Profile on the real surface, not from intuition: wrap the suspect node and read the live numbers. In node, `exec_js` carries `duration_ms` for free, surfaces your own timing and `process.memoryUsage()` on stdout, and lands the thrown-error `stack` on stderr -- read both channels (numbers on stdout, stack on stderr). In the browser, a body prefixed `capture\n<script>` auto-returns `{result, debug:{console, pageErrors, network, performance}}` with zero boilerplate. When the slow node is not obvious, sample it bottom-up: `exec_js` with `opts.profile:true` and the browser `profile\n<script>` prefix both return `{result, profile:{timeframe:{start_us,end_us,total_us,sample_count}, culprits:[{location,function,self_us,self_pct,hits}]}}` -- the worst-N `file:line` by self-time across init and code-execution, identical shape on both surfaces, so the culprit ranking points straight at the line to fix. Both also return `mem` (rss/heap/delta) and `wall_vs_cpu:{wall_us, offcpu_us}` -- the sampler sees only on-CPU JS, so a large `offcpu_us` means the time is going to IO, async wait, or the GPU, not the JS you can see; tune with `opts.sampleIntervalUs`/`opts.profileTopN` (cli) or `interval=`/`topN=` (browser). For the cheap non-profile path, `opts.mem:true` returns `{result, mem, wall_ms}` plus a structured `error:{name,message,stack}` on a throw -- read `error.name` directly instead of grepping the stderr stack; the default path (no `opts.mem`) stays byte-unchanged. The CPU sampler is structurally blind to GPU activity -- when wall greatly exceeds CPU self-time on a render/canvas/WebGL surface, the browser `trace\n<script>` prefix opens CDP Tracing and returns `trace:{wall_us, gpu_us, viz_us, cc_us, by_category}`, the wall-clock GPU-process time the profiler cannot show. Profile to LOCATE the slow/broken node, then eliminate hypotheses by live measurement. Verification is the same labour: run the real thing and witness the real output (the single mock-free `test.js`, the live page, the real service), never an automated unit/mock harness standing in for the real-services witness. Apparent tooling failure is part of this -- it is your mechanical self-recovery by elimination, never a question for the user.
 
 ## Memorize
 

package/gm-plugkit/package.json

@@ -1,6 +1,6 @@
 {
   "name": "gm-plugkit",
-  "version": "2.0.1632",
+  "version": "2.0.1634",
   "description": "Bootstrap and daemon-spawn tool for gm plugkit binary. Downloads the correct platform binary, verifies SHA256, and starts the spool watcher daemon. Includes plugkit-wasm-wrapper for WASM-based spool watching.",
   "main": "index.js",
   "bin": {

package/gm-plugkit/plugkit-wasm-wrapper.js

@@ -672,6 +672,18 @@ const AGGREGATE_CPU_PROFILE_SRC = `function aggregateCpuProfile(profile, topN) {
 }`;
 
 let execProfileSeq = 0;
+function sweepStaleProfileTmp() {
+  try {
+    const dir = os.tmpdir();
+    const cutoff = Date.now() - 3600000;
+    for (const name of fs.readdirSync(dir)) {
+      if (!/^gm-prof-\d+-\d+\.js$/.test(name)) continue;
+      const fp = path.join(dir, name);
+      try { if (fs.statSync(fp).mtimeMs < cutoff) fs.unlinkSync(fp); } catch (_) {}
+    }
+  } catch (_) {}
+}
+try { sweepStaleProfileTmp(); } catch (_) {}
 let _aggregateCpuProfileFn = null;
 function aggregateCpuProfile(profile, topN) {
   if (!_aggregateCpuProfileFn) {
@@ -861,6 +873,62 @@ function reapOrphanBrowserSessions(pw, cwd, claudeSessionId, reason) {
   }
 }
 
+const __openedSessionIds = new Set();
+function enumerateManagedChromiums(profileRootMarker) {
+  const marker = String(profileRootMarker || '').toLowerCase().replace(/\\/g, '/');
+  const out = [];
+  try {
+    if (process.platform === 'win32') {
+      const ps = `Get-CimInstance Win32_Process -Filter "Name='chrome.exe'" | Where-Object { $_.CommandLine -like '*--remote-debugging-port*' -and $_.CommandLine -like '*browser-profile*' -and $_.CommandLine -notlike '*--type=*' } | ForEach-Object { $_.ProcessId.ToString() + '|' + $_.CommandLine }`;
+      const r = spawnSync('powershell.exe', ['-NoProfile', '-NonInteractive', '-Command', ps], { encoding: 'utf-8', windowsHide: true, timeout: 10000 });
+      if (r.status === 0 && r.stdout) {
+        for (const line of r.stdout.split(/\r?\n/).filter(Boolean)) {
+          const bar = line.indexOf('|');
+          if (bar < 0) continue;
+          const pid = parseInt(line.slice(0, bar), 10);
+          const cmd = line.slice(bar + 1);
+          if (/--type=/.test(cmd)) continue;
+          if (Number.isFinite(pid) && cmd.toLowerCase().replace(/\\/g, '/').includes(marker)) out.push({ pid, cmd });
+        }
+      }
+    } else {
+      const r = spawnSync('ps', ['-eo', 'pid,command'], { encoding: 'utf-8', timeout: 10000 });
+      if (r.status === 0 && r.stdout) {
+        for (const line of r.stdout.split('\n').slice(1)) {
+          if (!/--remote-debugging-port/.test(line) || !/browser-profile/.test(line)) continue;
+          if (/--type=/.test(line)) continue;
+          const m = line.match(/^\s*(\d+)\s+(.+)$/);
+          if (!m) continue;
+          const pid = parseInt(m[1], 10);
+          if (Number.isFinite(pid) && m[2].toLowerCase().replace(/\\/g, '/').includes(marker)) out.push({ pid, cmd: m[2] });
+        }
+      }
+    }
+  } catch (_) {}
+  return out;
+}
+function reapOrphanChromiums(cwd, reason) {
+  try {
+    const root = browserRootDir(cwd);
+    const marker = path.join(root, '.gm', 'browser-profile').toLowerCase().replace(/\\/g, '/');
+    const procs = enumerateManagedChromiums(marker);
+    if (procs.length === 0) return { reaped: 0 };
+    const ports = readJsonFile(browserPortsFile(cwd), {});
+    const livePids = new Set();
+    for (const ent of Object.values(ports)) {
+      if (ent && Number.isFinite(ent.pid) && isProcessAliveSync(ent.pid)) livePids.add(ent.pid);
+    }
+    let reaped = 0;
+    for (const { pid } of procs) {
+      if (livePids.has(pid)) continue;
+      try { killPidQuiet(pid); reaped++; logEvent('plugkit', 'browser.os-orphan-reaped', { pid, reason: reason || 'sweep' }); } catch (_) {}
+    }
+    return { reaped };
+  } catch (_) {
+    return { reaped: 0 };
+  }
+}
+
 function resolveWindowsExeLocal(cmd) {
   if (process.platform !== 'win32') return cmd;
   try {
@@ -1295,6 +1363,10 @@ function getOrCreateBrowserSession(cwd, claudeSessionId, pw) {
   sessions[claudeSessionId] = [pwSessionId];
   writeJsonFile(portsFile, ports);
   writeJsonFile(sessionsFile, sessions);
+  if (!__openedSessionIds.has(claudeSessionId) && __openedSessionIds.size >= 1) {
+    logEvent('hook', 'deviation.browser-multi-session', { sid: claudeSessionId, already_open: Array.from(__openedSessionIds), reason: 'a 2nd distinct browser sessionId launched its own chromium this run -- reuse one session per run and close it when done' });
+  }
+  __openedSessionIds.add(claudeSessionId);
   logEvent('plugkit', 'browser.attached', { pwSessionId, pid: browserPid, port });
   return pwSessionId;
   } finally { releaseSpawnLock(); }
@@ -2032,6 +2104,18 @@ function makeHostFunctions(instanceRef) {
               + `  __session.disconnect();\n`
               + `})();\n`;
             cmd = process.execPath; args = ['-e', runnerCode];
+          } else if (opts.mem === true) {
+            const memRunner = `const { performance: __perf } = require('perf_hooks');\n`
+              + `(async () => {\n`
+              + `  const __mb = process.memoryUsage(); const __w0 = __perf.now();\n`
+              + `  let __r = null, __err = null;\n`
+              + `  try { __r = await (async () => {\n${code}\n})(); } catch (e) { __err = { name: e && e.name || 'Error', message: String(e && e.message || e), stack: String(e && e.stack || '') }; }\n`
+              + `  const __wallMs = Math.round((__perf.now() - __w0) * 1000) / 1000; const __ma = process.memoryUsage();\n`
+              + `  const __mem = { rss_mb: Math.round(__ma.rss/10485.76)/100, heapUsed_mb: Math.round(__ma.heapUsed/10485.76)/100, heapUsed_delta_mb: Math.round((__ma.heapUsed-__mb.heapUsed)/10485.76)/100, external_mb: Math.round(__ma.external/10485.76)/100 };\n`
+              + `  process.stdout.write('__GM_META__' + JSON.stringify({ result: __r === undefined ? null : __r, error: __err, mem: __mem, wall_ms: __wallMs }));\n`
+              + `  if (__err) process.exitCode = 1;\n`
+              + `})();\n`;
+            cmd = process.execPath; args = ['-e', memRunner];
           } else {
             cmd = process.execPath; args = ['-e', code];
           }
@@ -2041,8 +2125,12 @@ function makeHostFunctions(instanceRef) {
         else if (lang === 'deno') { cmd = 'deno'; args = ['eval', code]; }
         else { return writeWasmJson(instanceRef.value, { ok: false, error: `unsupported lang: ${lang}` }); }
         const __execT0 = Date.now();
-        const result = spawnSync(cmd, args, { encoding: 'utf-8', timeout: timeoutMs, cwd, env: process.env });
-        if (profileUserFile) { try { fs.unlinkSync(profileUserFile); } catch (_) {} }
+        let result;
+        try {
+          result = spawnSync(cmd, args, { encoding: 'utf-8', timeout: timeoutMs, cwd, env: process.env });
+        } finally {
+          if (profileUserFile) { try { fs.unlinkSync(profileUserFile); } catch (_) {} }
+        }
         if (wantProfile) {
           const raw = result.stdout || '';
           const idx = raw.indexOf('__GM_PROFILE__');
@@ -2063,6 +2151,24 @@ function makeHostFunctions(instanceRef) {
             wall_vs_cpu: parsed ? parsed.wall_vs_cpu : null,
           });
         }
+        if (opts.mem === true && isJsLang) {
+          const raw = result.stdout || '';
+          const idx = raw.indexOf('__GM_META__');
+          let meta = null;
+          if (idx >= 0) { try { meta = JSON.parse(raw.slice(idx + '__GM_META__'.length)); } catch (_) {} }
+          return writeWasmJson(instanceRef.value, {
+            ok: result.status === 0 && !!meta && !meta.error,
+            stdout: idx >= 0 ? raw.slice(0, idx) : raw,
+            stderr: result.stderr || '',
+            exit_code: result.status === null ? -1 : result.status,
+            timed_out: result.signal === 'SIGTERM',
+            duration_ms: Date.now() - __execT0,
+            result: meta ? meta.result : null,
+            mem: meta ? meta.mem : null,
+            wall_ms: meta ? meta.wall_ms : null,
+            ...(meta && meta.error ? { error: meta.error } : {}),
+          });
+        }
         return writeWasmJson(instanceRef.value, {
           ok: result.status === 0,
           stdout: result.stdout || '',
@@ -2141,6 +2247,7 @@ function makeHostFunctions(instanceRef) {
             stderr: '',
             exit_code: 0,
             session_id: pwSessionId,
+            hint: 'Reuse this same session for every browser dispatch this run (the spool sessionId selects it); a different sessionId opens its OWN chromium. Close it with `session close` when done -- the idle/orphan reaper is only a backstop.',
           });
         }
 
@@ -2198,6 +2305,22 @@ function makeHostFunctions(instanceRef) {
             evalBody = 'return {url: page.url(), title: await page.title()};';
           }
         }
+        let screenshotPath = null;
+        const shotMatch = evalBody.match(/^screenshot(?:=(\S+))?[ \t]*\n([\s\S]*)$/);
+        if (shotMatch) {
+          const witnessDir = path.join(browserRootDir(cwd), '.gm', 'witness');
+          try { fs.mkdirSync(witnessDir, { recursive: true }); } catch (_) {}
+          const reqName = shotMatch[1] ? path.basename(shotMatch[1]).replace(/[^A-Za-z0-9._-]/g, '_') : '';
+          const fname = (reqName && /\.png$/i.test(reqName)) ? reqName : `shot-${process.pid}-${execProfileSeq++}.png`;
+          screenshotPath = path.join(witnessDir, fname);
+          evalBody = shotMatch[2];
+        }
+        let domSelector = null;
+        const domMatch = evalBody.match(/^dom=(.+?)[ \t]*\n([\s\S]*)$/);
+        if (domMatch) {
+          domSelector = domMatch[1];
+          evalBody = domMatch[2] && domMatch[2].trim() ? domMatch[2] : 'return null;';
+        }
         const navTimeout = Math.min(timeoutMs, 60000);
         const gotoPrefix = startUrl
           ? `await page.goto(${JSON.stringify(startUrl)},{waitUntil:'load',timeout:${navTimeout}});\n`
@@ -2212,12 +2335,15 @@ function makeHostFunctions(instanceRef) {
           + `try{page.on('console',m=>{try{__logs.push({type:m.type(),text:m.text()});}catch(_){}});`
           + `page.on('pageerror',e=>{try{__errs.push({type:'pageerror',msg:String(e&&e.message||e)});}catch(_){}});`
           + `page.on('error',e=>{try{__errs.push({type:'uncaught',msg:String(e&&e.message||e),stack:String(e&&e.stack||'')});}catch(_){}});`
-          + `page.on('requestfinished',r=>{try{const t=r.timing();__net.push({url:String(r.url()).slice(0,120),dur_ms:Math.round(t.responseEnd),ttfb_ms:Math.round(t.responseStart)});}catch(_){}});`
+          + `page.on('requestfinished',r=>{try{const t=r.timing();let __st=0,__sz=0;try{__st=(r.response()&&r.response().status())||0;}catch(_){}__net.push({url:String(r.url()).slice(0,120),method:r.method(),status:__st,dur_ms:Math.round(t.responseEnd),ttfb_ms:Math.round(t.responseStart)});}catch(_){}});`
           + `page.on('requestfailed',r=>{try{const err=r.failure();__errs.push({type:'fetch',msg:String(err&&err.errorText||'request failed'),url:String(r.url()).slice(0,120)});}catch(_){}});`
           + `page.evaluateOnNewDocument(()=>{window.__gmErrors=[];window.onerror=(msg,src,line,col,err)=>{try{window.__gmErrors.push({type:'error',msg:String(msg),src:String(src).slice(0,80),line,col,stack:String(err&&err.stack||'')});}catch(_){};return false;};window.onunhandledrejection=(e)=>{try{window.__gmErrors.push({type:'unhandledRejection',msg:String(e.reason&&e.reason.message||e.reason),stack:String(e.reason&&e.reason.stack||'')});}catch(_){}};});`
           + `}catch(_){}\n`;
-        const perfRead = `let __perf=null;try{__perf=await page.evaluate(async()=>{const n=performance.getEntriesByType('navigation')[0];const paints={};for(const p of performance.getEntriesByType('paint')){paints[p.name]=Math.round(p.startTime);}let lcp=0;try{const le=performance.getEntriesByType('largest-contentful-paint');if(le.length)lcp=Math.round(le[le.length-1].startTime);}catch(_){}let cls=0;try{for(const ls of performance.getEntriesByType('layout-shift')){if(!ls.hadRecentInput)cls+=ls.value;}}catch(_){}let longtasks=0;try{longtasks=performance.getEntriesByType('longtask').length;}catch(_){}const fps=await new Promise(res=>{let f=0;const s=performance.now();function tick(){f++;if(performance.now()-s>=500)return res(Math.round(f/((performance.now()-s)/1000)));requestAnimationFrame(tick);}requestAnimationFrame(tick);});return{load_ms:n?Math.round(n.loadEventEnd||0):0,dcl_ms:n?Math.round(n.domContentLoadedEventEnd||0):0,resources:performance.getEntriesByType('resource').length,now:Math.round(performance.now()),first_paint_ms:paints['first-paint']||0,first_contentful_paint_ms:paints['first-contentful-paint']||0,largest_contentful_paint_ms:lcp,cumulative_layout_shift:Math.round(cls*1000)/1000,longtasks,fps};});}catch(_){}\n`;
+        const perfRead = `let __perf=null;try{__perf=await page.evaluate(async()=>{const n=performance.getEntriesByType('navigation')[0];const paints={};for(const p of performance.getEntriesByType('paint')){paints[p.name]=Math.round(p.startTime);}let lcp=0;try{const le=performance.getEntriesByType('largest-contentful-paint');if(le.length)lcp=Math.round(le[le.length-1].startTime);}catch(_){}let cls=0;try{for(const ls of performance.getEntriesByType('layout-shift')){if(!ls.hadRecentInput)cls+=ls.value;}}catch(_){}let longtasks=0;try{longtasks=performance.getEntriesByType('longtask').length;}catch(_){}let heapU=0,heapT=0;try{if(performance.memory){heapU=Math.round(performance.memory.usedJSHeapSize/10485.76)/100;heapT=Math.round(performance.memory.totalJSHeapSize/10485.76)/100;}}catch(_){}const fps=await new Promise(res=>{let f=0;const s=performance.now();function tick(){f++;if(performance.now()-s>=500)return res(Math.round(f/((performance.now()-s)/1000)));requestAnimationFrame(tick);}requestAnimationFrame(tick);});return{load_ms:n?Math.round(n.loadEventEnd||0):0,dcl_ms:n?Math.round(n.domContentLoadedEventEnd||0):0,resources:performance.getEntriesByType('resource').length,now:Math.round(performance.now()),first_paint_ms:paints['first-paint']||0,first_contentful_paint_ms:paints['first-contentful-paint']||0,largest_contentful_paint_ms:lcp,cumulative_layout_shift:Math.round(cls*1000)/1000,longtasks,fps,heap_used_mb:heapU,heap_total_mb:heapT};});}catch(_){}\n`;
         const blankProbe = startUrl ? '' : `try{const __u=page.url();if(__u==='about:blank'||__u===''){console.error('__GM_BLANK__');}}catch(_){}\n`;
+        const netFmt = `__net.slice().sort((a,b)=>(b.dur_ms||0)-(a.dur_ms||0)).slice(0,30)`;
+        const consoleFmt = `(__logs.length>50?[...__logs.slice(0,50),{type:'meta',text:'... '+(__logs.length-50)+' more console entries dropped'}]:__logs)`;
+        const emitResult = `try{console.log('__GM_RESULT__'+JSON.stringify(__RET===undefined?null:__RET));}catch(__se){console.log('__GM_RESULT__'+JSON.stringify({__unserializable:String(__se&&__se.message||__se)}));}\n`;
         if (modeMatch && modeMatch[1] === 'profile') {
           const userScript = modeMatch[3];
           const intervalUs = sampleIntervalUs;
@@ -2236,7 +2362,8 @@ function makeHostFunctions(instanceRef) {
             + `const __cpuUs=__agg.timeframe?__agg.timeframe.total_us:0;\n`
             + `const __wallVsCpu={wall_us:__wallUs,cpu_self_us:__cpuUs,offcpu_us:Math.max(0,__wallUs-__cpuUs),note:'offcpu_us = wall minus on-CPU JS self time = GPU/compositor/raster/IO/idle the CPU sampler is blind to; use trace mode to attribute GPU activity'};\n`
             + `const __allErrors=[...__errs,...__wmErrors];\n`
-            + `return {result:__result,profile:__agg,profile_error:__profileError,wall_vs_cpu:__wallVsCpu,debug:{console:__logs,pageErrors:__allErrors,network:__net.slice(0,30),performance:__perf}};`;
+            + `const __RET={result:__result,profile:__agg,profile_error:__profileError,wall_vs_cpu:__wallVsCpu,debug:{console:${consoleFmt},pageErrors:__allErrors,network:${netFmt},performance:__perf}};\n`
+            + emitResult + `return __RET;`;
         } else if (modeMatch && modeMatch[1] === 'trace') {
           const userScript = modeMatch[3];
           evalBody = debugSetup
@@ -2255,7 +2382,8 @@ function makeHostFunctions(instanceRef) {
             + `const __topCats=Object.entries(__byCat).sort((a,b)=>b[1]-a[1]).slice(0,15).map(([cat,us])=>({cat,wall_us:us}));\n`
             + `const __spanUs=(isFinite(__minTs)&&__maxTs>0)?(__maxTs-__minTs):0;\n`
             + `const __allErrors=[...__errs,...__wmErrors];\n`
-            + `return {result:__result,trace:{wall_us:__wallUs,trace_span_us:__spanUs,event_count:__traceEvents.length,complete:__traceComplete,gpu_us:__gpuUs,viz_us:__vizUs,cc_us:__ccUs,raster_us:__rasterUs,offcpu_note:'gpu_us/viz_us/cc_us are wall-clock GPU-process activity (compositor/raster/draw) captured via CDP Tracing -- the CPU sampler cannot see these',by_category:__topCats},trace_error:__traceError,debug:{console:__logs,pageErrors:__allErrors,network:__net.slice(0,30),performance:__perf}};`;
+            + `const __RET={result:__result,trace:{wall_us:__wallUs,trace_span_us:__spanUs,event_count:__traceEvents.length,complete:__traceComplete,gpu_us:__gpuUs,viz_us:__vizUs,cc_us:__ccUs,raster_us:__rasterUs,offcpu_note:'gpu_us/viz_us/cc_us are wall-clock GPU-process activity (compositor/raster/draw) captured via CDP Tracing -- the CPU sampler cannot see these',by_category:__topCats},trace_error:__traceError,debug:{console:${consoleFmt},pageErrors:__allErrors,network:${netFmt},performance:__perf}};\n`
+            + emitResult + `return __RET;`;
         } else if (modeMatch && modeMatch[1] === 'capture') {
           const userScript = modeMatch[3];
           evalBody = debugSetup
@@ -2263,11 +2391,22 @@ function makeHostFunctions(instanceRef) {
             + `const __wmErrors=await page.evaluate(()=>window.__gmErrors||[]);\n`
             + perfRead
             + `const __allErrors=[...__errs,...__wmErrors];\n`
-            + `return {result:__result,debug:{console:__logs,pageErrors:__allErrors,network:__net.slice(0,30),performance:__perf}};`;
+            + `const __RET={result:__result,debug:{console:${consoleFmt},pageErrors:__allErrors,network:${netFmt},performance:__perf}};\n`
+            + emitResult + `return __RET;`;
+        } else if (screenshotPath) {
+          evalBody = `const __result = await (async () => {\n${blankProbe}${gotoPrefix}try{${evalBody}}catch(e){throw e;}\n})();\n`
+            + `let __shotErr=null;try{await page.screenshot({path:${JSON.stringify(screenshotPath)},fullPage:false});}catch(e){__shotErr=String(e&&e.message||e);}\n`
+            + `const __RET={result:__result,screenshot_path:${JSON.stringify(screenshotPath)},screenshot_error:__shotErr};\n`
+            + emitResult + `return __RET;`;
+        } else if (domSelector) {
+          evalBody = `${blankProbe}${gotoPrefix}let __RET;try{__RET=await page.evaluate((sel)=>{const out=[];const els=document.querySelectorAll(sel);for(let i=0;i<Math.min(els.length,20);i++){const e=els[i];const r=e.getBoundingClientRect();const attrs={};for(const a of e.attributes)attrs[a.name]=String(a.value).slice(0,120);out.push({tag:e.tagName.toLowerCase(),text:(e.textContent||'').trim().slice(0,200),attrs,visible:!!(r.width&&r.height),rect:{x:Math.round(r.x),y:Math.round(r.y),w:Math.round(r.width),h:Math.round(r.height)}});}return{selector:sel,match_count:els.length,elements:out};},${JSON.stringify(domSelector)});}catch(e){__RET={selector:${JSON.stringify(domSelector)},error:String(e&&e.message||e),match_count:0,elements:[]};}\n`
+            + emitResult + `return __RET;`;
         } else if (startUrl) {
-          evalBody = `${gotoPrefix}${evalBody}`;
+          evalBody = `${gotoPrefix}const __RET=await (async()=>{${evalBody}})();\n` + emitResult + `return __RET;`;
         } else if (blankProbe) {
-          evalBody = `${blankProbe}${evalBody}`;
+          evalBody = `${blankProbe}const __RET=await (async()=>{${evalBody}})();\n` + emitResult + `return __RET;`;
+        } else {
+          evalBody = `const __RET=await (async()=>{${evalBody}})();\n` + emitResult + `return __RET;`;
         }
         const outerTimeoutMs = Math.min(timeoutMs + 6000, 126000);
         const r = runBrowserRunner(pw, ['-s', pwSessionId, '--timeout', String(timeoutMs), '-e', evalBody], outerTimeoutMs, cwd, sessionId);
@@ -2277,15 +2416,30 @@ function makeHostFunctions(instanceRef) {
         }
         const rawStderr = r.stderr || '';
         const landedOnBlank = !startUrl && rawStderr.includes('__GM_BLANK__');
+        let rawStdout = r.stdout || '';
+        let parsedResult;
+        let resultParsed = false;
+        const resIdx = rawStdout.lastIndexOf('__GM_RESULT__');
+        if (resIdx >= 0) {
+          const tail = rawStdout.slice(resIdx + '__GM_RESULT__'.length);
+          const nl = tail.indexOf('\n');
+          const jsonStr = nl >= 0 ? tail.slice(0, nl) : tail;
+          try { parsedResult = JSON.parse(jsonStr); resultParsed = true; } catch (_) {}
+          rawStdout = (rawStdout.slice(0, resIdx) + (nl >= 0 ? tail.slice(nl + 1) : '')).replace(/\n+$/, '\n');
+        }
         const envelope = {
           ok,
-          stdout: scrubBrowserRunnerText(r.stdout || ''),
+          stdout: scrubBrowserRunnerText(rawStdout),
           stderr: scrubBrowserRunnerText(rawStderr.replace(/^__GM_BLANK__\r?\n?/gm, '')),
           exit_code: r.status === null ? -1 : r.status,
           session_id: pwSessionId,
           timeout_ms_used: timeoutMs,
         };
+        if (resultParsed) envelope.result = parsedResult;
         envelope.navigation_requested = !!startUrl;
+        if (__openedSessionIds.size > 1) {
+          envelope.multi_session_warning = `${__openedSessionIds.size} distinct browser sessions opened this run, each its own chromium -- reuse ONE sessionId per run and 'session close' it when done to avoid leaking browsers.`;
+        }
         if (landedOnBlank) {
           envelope.landed_on_blank = true;
           envelope.hint = "page is about:blank: this dispatch did not navigate, so the expression evaluated against an empty page. Prefix the body with 'url=<target>' (or send a bare 'https://...' URL) to open the page you want before evaluating.";
@@ -2398,6 +2552,7 @@ async function runSpoolWatcher(instance, spoolDir) {
   } catch (_) {}
 
   try { reapOrphanBrowserSessions(findBrowserRunner(), process.cwd(), process.env.CLAUDE_SESSION_ID || 'claude-loop-iter', 'watcher-boot'); } catch (_) {}
+  try { reapOrphanChromiums(process.cwd(), 'watcher-boot'); } catch (_) {}
 
 
   const LOCK_PATH = path.join(spoolDir, '.watcher.lock');
@@ -3034,7 +3189,7 @@ async function runSpoolWatcher(instance, spoolDir) {
     }
   }, 60_000);
 
-  const BROWSER_IDLE_LIMIT_MS = parseInt(process.env.PLUGKIT_BROWSER_IDLE_LIMIT_MS, 10) || 10 * 60 * 1000;
+  const BROWSER_IDLE_LIMIT_MS = parseInt(process.env.PLUGKIT_BROWSER_IDLE_LIMIT_MS, 10) || 3 * 60 * 1000;
   setInterval(() => {
     try {
       const portsFile = browserPortsFile(process.cwd());
@@ -3078,6 +3233,7 @@ async function runSpoolWatcher(instance, spoolDir) {
         try { writeJsonFile(sessionsFile, sessions); } catch (_) {}
       }
       try { reapOrphanBrowserSessions(findBrowserRunner(), process.cwd(), process.env.CLAUDE_SESSION_ID || 'claude-loop-iter', 'idle-sweep'); } catch (_) {}
+      try { reapOrphanChromiums(process.cwd(), 'idle-sweep'); } catch (_) {}
     } catch (e) {
       console.error(`[browser-idle] error: ${e.message}`);
     }

package/gm.json

@@ -1,6 +1,6 @@
 {
   "name": "gm",
-  "version": "2.0.1632",
+  "version": "2.0.1634",
   "description": "Spool-dispatch orchestration engine with unified state machine, skills, and automated git enforcement",
   "author": "AnEntrypoint",
   "license": "MIT",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "gm-skill",
-  "version": "2.0.1632",
+  "version": "2.0.1634",
   "description": "Canonical universal harness — AI-native software engineering via skill-driven orchestration; bootstraps plugkit for task execution and session isolation. Install in any AI coding agent host.",
   "author": "AnEntrypoint",
   "license": "MIT",