fix(newos4): coreutils flags + loop-guard arg key, both found by the agent's live run (v4.6.49)

Agent session surfaced two real defects:

1. The A4 prompt told it to use grep -n / sed -n 'N,Mp' / wc -l / head -N — none of those flags
   parsed: grep treated '-n' as the pattern (6/6 variants failed), wc read '-l' as a filename,
   head looked for a file named '-2', sed had no range parser. Fixed the tools, not just the text:
   grep now accepts -n/-i/-c/-v/-l (line numbers were already printed), wc accepts -l/-w/-c,
   head accepts -N and -n N, sed accepts Np and N,Mp, tail no longer prints a blank line for a
   trailing newline. Prompt re-ranked to the working forms.

2. Loop guard keyed calls by tool name plus the raw arguments field. When arguments was already
   an object (parsed upstream), the key became 'run_command:[object Object]', collapsing distinct
   batched commands into one signature and producing false "called 3x" warnings. Now the key is
   JSON.stringify(arguments) whenever it is not a string (3 sites: sig, _sigR, _sigD).

Verified live v4.6.49: grep -n/-c, wc -l, head -2, sed -n '2,3p', tail -n 1 all correct.

1 files changed, 12 insertions, 5 deletions

diff --git a/modules/core/module-aiassist.js b/modules/core/module-aiassist.js
index 8d1fc9d..f69494e 100644
--- a/modules/core/module-aiassist.js
+++ b/modules/core/module-aiassist.js
@@ -122,7 +122,7 @@ buildSystemPrompt() {
                 } catch (e) {}
                 base += '\n\n[File reads, tools, verification]\n' +
                     '- read_file returns 8000 chars by default (limit up to 16000). If the reply says truncated=true, call read_file again with offset=<next_offset> before concluding anything about the file.\n' +
-                    '- Prefer targeted reads over whole files: run_command with grep -n "<pattern>" <file>, sed -n \'10,40p\' <file>, head -40 <file>, tail -40 <file>, wc -l <file>.\n' +
+                    '- Prefer targeted reads over whole files: run_command with grep "<pattern>" <file> (prints line: text), sed -n \'10,40p\' <file>, head -40 <file>, tail -n 40 <file>, wc -l <file>. Flags are tolerated, but plain `cat <file>` and read_file offset/limit always work.\n' +
                     '- sed without -i is read-only and allowed; sed -i, rm, mv and writes need user confirmation.\n' +
                     '- After ANY write or edit: re-read the changed lines (cat or sed -n) and run a check that exists in THIS terminal — `test` for the suite, `grep -n` to confirm the new line, `wc -l` for size. There is no shell, no node and no curl here.\n' +
                     '- If a tool fails twice, change approach (different tool, smaller scope) instead of retrying the same call.';
@@ -1540,7 +1540,14 @@ buildSystemPrompt() {
                         const now = Date.now();
                         const blocked = [];
                         for (const tc of res.tool_calls) {
-                            const sig = tc.function.name + ':' + (tc.function.arguments || '');
+                            // 2026-10-21: arguments can be an object (already parsed upstream).
+                            // `+ object` produced "name:[object Object]", collapsing every call to
+                            // the same tool into one key → false "called 3x" loop warnings when the
+                            // model legitimately batched distinct commands.
+                            const _argKey = (typeof tc.function.arguments === 'string')
+                                ? tc.function.arguments
+                                : JSON.stringify(tc.function.arguments || {});
+                            const sig = tc.function.name + ':' + _argKey;
                             let entry = _seenCalls.get(sig);
                             if (!entry || (now - entry.ts) > _loopWindow) {
                                 entry = { count: 0, ts: now, resultHash: null, blocked: false, lastResult: null, epoch: this._toolEpoch };
@@ -1555,7 +1562,7 @@ buildSystemPrompt() {
                             entry.ts = now;
                             _seenCalls.set(sig, entry);
 
-                            const argPreview = String(tc.function.arguments || '')
+                            const argPreview = String(_argKey || '')
                                 .replace(/(token|key|secret|password|bearer)[^,}]*/ig, '$1:[redacted]')
                                 .slice(0, 60);
                             if (entry.blocked || entry.count >= _blockAt) {
@@ -1668,7 +1675,7 @@ buildSystemPrompt() {
                                 this.terminal.print(`📊 Result (${t1}s): ${obs}`, 'info');
                             }
                             try {
-                                const _sigR = fn.name + ':' + (fn.arguments || '');
+                                const _sigR = fn.name + ':' + ((typeof fn.arguments === 'string') ? fn.arguments : JSON.stringify(fn.arguments || {}));
                                 const _eR = _seenCalls.get(_sigR);
                                 if (_eR) {
                                     const _rh = _sigHash(toolResult);
@@ -1743,7 +1750,7 @@ buildSystemPrompt() {
                                         this.terminal.print(`📊 Result (${t1}s): ${dObs}`, 'info');
                                     }
                                     try {
-                                        const _sigD = fn.name + ':' + (fn.arguments || '');
+                                        const _sigD = fn.name + ':' + ((typeof fn.arguments === 'string') ? fn.arguments : JSON.stringify(fn.arguments || {}));
                                         const _eD = _seenCalls.get(_sigD);
                                         if (_eD) {
                                             const _rhD = _sigHash(toolResult);