fix: address CodeRabbit review comments

- Fix runCalls: invoke callables before Promise.allSettled
- Add parseIntSafe: validate CLI options are positive integers
- Update SKILL.md with --max-tokens, --timeout, --max-retries, --sequential

Addresses CodeRabbit comments on PR #95.

Author: 23r2efewvcs

skills/llm-council/SKILL.md

@@ -43,10 +43,20 @@ Default model rosters per provider live in `scripts/council.js` and can be overr
 
 ```
 node $SKILL_ROOT/scripts/council.js run "<query>" [--models id1,id2,id3] [--chairman id] [--provider <name>] [--wiki <slug>]
+                        [--max-tokens N] [--timeout ms] [--max-retries N] [--sequential]
 node $SKILL_ROOT/scripts/council.js providers
 node $SKILL_ROOT/scripts/council.js show <session-id>
 ```
 
+### Runtime options
+
+| Flag | Default | Description |
+|------|---------|-------------|
+| `--max-tokens` | 4000 | Max output tokens per model call. Bump to 16000+ for reasoning models. |
+| `--timeout` | 120000 | HTTP request timeout in ms. Bump to 300000+ for slow endpoints (NVIDIA NIM). |
+| `--max-retries` | 1 | Retry count on 429/5xx and connection errors. Exponential backoff (2s, 4s, 8s...). |
+| `--sequential` | false | Run model calls one at a time instead of in parallel. Use when free endpoints reject concurrent requests. |
+
 `--wiki <slug>` writes the full transcript to `<wiki>/derived/council/<session-id>.md` and registers it via `wiki-cli.js page` so it shows in FTS5 search.
 
 ## Output

skills/llm-council/scripts/council.js

@@ -209,15 +209,23 @@ async function cmdRun(args) {
   const provider = PROVIDERS[providerName];
   if (!provider.baseUrl) { console.error(`provider ${providerName} requires LLM_COUNCIL_BASE_URL`); process.exit(2); }
 
-  if (args['max-tokens']) RUN_OPTS.max_tokens = parseInt(args['max-tokens'], 10);
-  if (args.timeout) RUN_OPTS.timeout_ms = parseInt(args.timeout, 10);
-  if (args['max-retries']) RUN_OPTS.max_retries = parseInt(args['max-retries'], 10);
+  function parseIntSafe(val, name) {
+    const n = parseInt(val, 10);
+    if (isNaN(n) || n <= 0 || n !== Math.floor(n)) {
+      console.error(`Invalid --${name}: ${val} (must be a positive integer)`);
+      process.exit(2);
+    }
+    return n;
+  }
+  if (args['max-tokens']) RUN_OPTS.max_tokens = parseIntSafe(args['max-tokens'], 'max-tokens');
+  if (args.timeout) RUN_OPTS.timeout_ms = parseIntSafe(args.timeout, 'timeout');
+  if (args['max-retries']) RUN_OPTS.max_retries = parseIntSafe(args['max-retries'], 'max-retries');
   if (args.sequential) RUN_OPTS.sequential = true;
 
   // Run a list of async callables either in parallel (default) or sequentially.
   // Sequential mode avoids concurrent-request limits on free NIM/OpenRouter endpoints.
   async function runCalls(callables) {
-    if (!RUN_OPTS.sequential) return Promise.allSettled(callables);
+    if (!RUN_OPTS.sequential) return Promise.allSettled(callables.map(fn => fn()));
     const results = [];
     for (const fn of callables) {
       try { results.push({ status: 'fulfilled', value: await fn() }); }