majlis 0.3.1 → 0.3.2
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/cli.js +24 -5
- package/package.json +1 -1
package/dist/cli.js
CHANGED
|
@@ -522,13 +522,29 @@ Before building:
|
|
|
522
522
|
3. Check docs/classification/ for problem taxonomy
|
|
523
523
|
4. Check docs/experiments/ for prior work
|
|
524
524
|
|
|
525
|
-
|
|
526
|
-
|
|
527
|
-
|
|
525
|
+
## Scope Constraint (CRITICAL)
|
|
526
|
+
|
|
527
|
+
You get ONE attempt per cycle. Your job is:
|
|
528
|
+
1. Read and diagnose \u2014 understand the problem thoroughly
|
|
529
|
+
2. Form ONE hypothesis about what to fix
|
|
530
|
+
3. Implement ONE focused change (not a multi-step debug session)
|
|
531
|
+
4. Run the benchmark ONCE to see the result
|
|
532
|
+
5. Document what you did, what happened, and what you think the next step should be
|
|
533
|
+
6. STOP
|
|
534
|
+
|
|
535
|
+
Do NOT iterate. Do NOT try multiple approaches. Do NOT debug your own fix.
|
|
536
|
+
If your change doesn't work, document why and let the cycle continue \u2014
|
|
537
|
+
the adversary, critic, and verifier will help diagnose what went wrong.
|
|
538
|
+
The cycle will come back to you with their insights.
|
|
539
|
+
|
|
540
|
+
If you find yourself wanting to "try one more thing," that's the signal to stop
|
|
541
|
+
and write up what you learned. The other agents exist precisely for this reason.
|
|
542
|
+
|
|
543
|
+
## During building:
|
|
528
544
|
- Tag EVERY decision: proof / test / strong-consensus / consensus / analogy / judgment
|
|
529
545
|
- When making judgment-level decisions, state: "This is judgment \u2014 reasoning without precedent"
|
|
530
546
|
- Run baseline metrics BEFORE making changes
|
|
531
|
-
- Run comparison metrics AFTER making changes
|
|
547
|
+
- Run comparison metrics AFTER making changes (once)
|
|
532
548
|
|
|
533
549
|
You may NOT verify your own work or mark your own decisions as proven.
|
|
534
550
|
Output your decisions in structured format so they can be recorded in the database.
|
|
@@ -2323,7 +2339,7 @@ async function runQuery(opts) {
|
|
|
2323
2339
|
cwd: opts.cwd,
|
|
2324
2340
|
permissionMode: "bypassPermissions",
|
|
2325
2341
|
allowDangerouslySkipPermissions: true,
|
|
2326
|
-
maxTurns:
|
|
2342
|
+
maxTurns: 30,
|
|
2327
2343
|
persistSession: false,
|
|
2328
2344
|
settingSources: ["project"]
|
|
2329
2345
|
}
|
|
@@ -2363,6 +2379,9 @@ async function runQuery(opts) {
|
|
|
2363
2379
|
} else if (message.type === "result") {
|
|
2364
2380
|
if (message.subtype === "success") {
|
|
2365
2381
|
costUsd = message.total_cost_usd;
|
|
2382
|
+
} else if (message.subtype === "error_max_turns") {
|
|
2383
|
+
costUsd = "total_cost_usd" in message ? message.total_cost_usd : 0;
|
|
2384
|
+
console.warn(`[majlis] Agent hit max turns (${turnCount}). Returning partial output.`);
|
|
2366
2385
|
} else {
|
|
2367
2386
|
const errors = "errors" in message ? message.errors?.join("; ") ?? "Unknown error" : "Unknown error";
|
|
2368
2387
|
throw new Error(`Agent query failed (${message.subtype}): ${errors}`);
|