bullswarm 0.10.7 → 0.10.9

package/AGENTS.md

@@ -37,7 +37,7 @@ content. Published as `bullswarm` on npm.
 ## Development
 
 ```bash
-npm test            # 261 tests, no network needed (meters read from cache)
+npm test            # full suite, no network needed (meters read from cache)
 node bin/bullswarm.js doctor --json   # readiness report
 node bin/bullswarm.js workflow list   # discover workflows
 node bin/bullswarm.js workflow runs   # ongoing workflow instances

package/CHANGELOG.md

@@ -1,5 +1,42 @@
 # bullswarm changelog
 
+## 0.10.9 — planner self-correction and honest silence
+
+- An invalid or non-JSON orchestrator decision no longer fails the run. The
+  runtime feeds the exact validator issues, the rejected proposal, and a
+  response excerpt back to the same orchestrator thread for bounded corrective
+  turns (`settings.maxPlannerCorrections`, default 2, emits
+  `decision.correction_requested`), then benches that pool as orchestrator for
+  the rest of the run and escalates to one other eligible pool
+  (`decision.orchestrator_escalated`), and only then settles on a qualified
+  `completed_with_concerns`/`blocked` outcome. The decide action's ledger status
+  now agrees with that outcome (`failed_retryable` while correcting,
+  `failed_terminal` on exhaustion) instead of reporting `succeeded`.
+- The planner prompt now shows complete `run`, `fanout` (`items` +
+  `stepTemplate.prompt` with `{{item}}`), and `verify` (`review`) skeletons, so
+  the first proposal can match what `decision.js` validates.
+- `workflow watch` separates two silences: `quiet` counts durable workflow
+  events, and a new `agent output … ago` figure (JSONL `transportQuietForSec`)
+  counts raw output from live agents, so a thinking agent and a dead one look
+  different on the same heartbeat line.
+- Removed the dead thin-leaf help renderer left over from the help unification
+  and stopped hard-coding the test count in AGENTS.md.
+
+## 0.10.8 — quieter monitoring and resilient orchestration
+
+- Made `workflow watch` aggregate low-level activity into compact interval
+  heartbeats by default, with event/action deltas, quiet duration, prompt
+  semantic transitions, terminal result handoff, and `--verbose` drill-down.
+- Implemented the documented top-level `run --prompt` form and standardized
+  usage errors as exit 2 across run, workflow drafts/runs, and strategy paths.
+- Added explicit goal-resume orchestrator pin/unpin behavior and strengthened
+  the orchestrator as a control-plane-only decision thread.
+- Recognize Claude's exact `Failed to authenticate` response as an auth failure
+  and migrate the connector signature additively without replacing local
+  connector customization.
+- Expanded non-network CLI, health, release, watch, resume, auth, help, and
+  documentation coverage. The full suite now contains 271 tests.
+
 ## 0.10.7 — clearer orchestration overview
 
 - Replaced the orchestrator trace dump with a summary-first view showing what

package/README.md

@@ -67,6 +67,7 @@ bullswarm setup    # re-run or repair
 bullswarm pools    # meter state, pace position, quarantine status
 bullswarm strategy refresh --apply --yes  # approve capability-aware tier autopilot
 bullswarm run --lane analyze --add-dir ~/some-repo --task-file /tmp/t.md --json
+bullswarm run --lane analyze --add-dir ~/some-repo --prompt "Inspect the parser" --json
 bullswarm workflow goal "Fix the failing tests and verify the change" --cwd ~/some-repo
 bullswarm health   # re-judge saved outputs; catch gate failures
 ```
@@ -84,6 +85,17 @@ bullswarm health   # re-judge saved outputs; catch gate failures
 | `doctor` | Machine-readable readiness report; self-heals on first call |
 | `workflow` | Start an autonomous goal, or run / validate / draft / inspect explicit workflows and their live instances. |
 
+Discover and validate workflow definitions without executing them:
+
+```bash
+bullswarm workflow list
+bullswarm workflow list --json
+bullswarm workflow validate workflows/my-workflow.json
+```
+
+`workflow goal --request <path>` and `--run-id <id>` are internal detached-runner
+resume plumbing. Normal callers should provide a goal or use `--resume <shortId|runId>`.
+
 ## Model strategy and invocation telemetry
 
 Bullswarm can inventory the models exposed by installed agent CLIs and combine
@@ -260,16 +272,20 @@ auditing completed runs.
 
 ### Live workflow dashboard
 
-For ordinary observation, use the non-interactive watcher. It prints only when
-the phase, step, agent action, routing, or status changes, plus a 60-second
-heartbeat while otherwise quiet. Terminal output includes every attempt's
-agent/model, elapsed time, outcome, and tokens, so a slow test is distinguishable
-from a stalled process without writing a polling script.
+For ordinary observation, use the non-interactive watcher. Human output is one
+compact aggregate line per semantic change and a 60-second heartbeat while
+otherwise quiet. Each line reports status/location, events and agent actions
+captured since the preceding sample, and quiet duration. It does not repeat
+command or response excerpts. Use `--verbose` for the detailed per-agent and
+last-action view. Compact terminal output reports the overall attempt count and
+elapsed time; `--verbose` includes every attempt's agent/model, outcome, and
+tokens. This keeps agent monitoring cheap while retaining a drill-down path.
 
 ```bash
 bullswarm workflow watch <shortId>
 bullswarm workflow watch <shortId> --jsonl       # automation-friendly stream
 bullswarm workflow watch <shortId> --once        # one current/terminal snapshot
+bullswarm workflow watch <shortId> --verbose     # detailed agent/action view
 ```
 
 `workflow tui` is the interactive, Claude-style `/workflows` view. For an

package/connectors/claude-code.json

@@ -18,6 +18,7 @@
   "authSignatures": [
     "unauthorized",
     "authentication failed",
+    "failed to authenticate",
     "credit balance",
     "not logged in",
     "please run /login"

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "bullswarm",
-  "version": "0.10.7",
+  "version": "0.10.9",
   "description": "Route work across coding-agent CLI subscriptions — paced by live quota meters, verified by content, never trusting exit codes.",
   "type": "module",
   "bin": {

package/skill/SKILL.md

@@ -59,6 +59,10 @@ bullswarm run --lane <analyze|build|chore> \
   --add-dir <abs/path/to/repo> \
   --task-file <abs/path/to/task.md> \
   --json
+
+# Equivalent explicit prompt form (also accepts trailing task text)
+bullswarm run --lane analyze --add-dir <abs/path/to/repo> \
+  --prompt "Inspect and verify the parser" --json
 ```
 
 The verdict shape:
@@ -202,8 +206,15 @@ bullswarm workflow runs --historical --since yesterday --until today
 bullswarm workflow runs show <shortId>        # state + report + summary
 bullswarm workflow runs result <shortId> --json # stable delivery for the caller
 bullswarm workflow runs delete <shortId> --yes
+
+bullswarm workflow list
+bullswarm workflow validate workflows/demo.json
 ```
 
+`workflow goal --request <path>` and `--run-id <id>` are internal detached-runner
+plumbing. Ordinary callers should provide a goal or resume with
+`--resume <shortId|runId>`.
+
 Historical time ranges filter the workflow's initiation time (`startedAt`),
 not completion time. `--since` is inclusive and `--until` is exclusive, with
 `--from`/`--to` and `--started-after`/`--started-before` aliases. Bounds accept
@@ -279,9 +290,13 @@ reason, and child-process termination evidence. `workflow tui` displays the
 same information interactively. `workflow events` supports replay after a
 monotonic sequence cursor.
 
-Prefer `workflow watch` for ordinary monitoring; it prints semantic changes and
-a periodic heartbeat, then a per-attempt timing/token breakdown at terminal
-status. `workflow steer` is an optional durable queue for autonomous workflows:
+Prefer `workflow watch` for ordinary monitoring; default human text is a compact
+aggregate heartbeat with status/location, interval event and agent-action counts,
+and quiet duration. It does not repeat command or response excerpts. Add
+`--verbose` for the detailed agent and last-action view. `--jsonl` keeps the
+machine-readable snapshot stream. Compact terminal text includes overall timing
+and the exact `workflow runs result <shortId> --json` next command; `--verbose`
+adds per-attempt timing and token details. `workflow steer` is an optional durable queue for autonomous workflows:
 it never changes the active worker and is delivered only to the next
 not-yet-started decision gate. Steering cannot expand the original authority,
 weaken verification, or bypass proposal validation.
@@ -320,6 +335,14 @@ execute -> observe -> decide -> validate proposal -> append -> execute -> observ
 Allowed planner decisions are `proceed`, `complete`, `needs_more_work`,
 `retry`, `escalate`, `wait_for_approval`, and `stop`. Expansion decisions must
 contain bounded actions; malformed or over-budget output executes nothing.
+A malformed or non-JSON decision does not fail the run: the runtime returns
+the exact validator issues to the same orchestrator thread for up to
+`settings.maxPlannerCorrections` (default 2) corrective turns
+(`decision.correction_requested` events), then benches that pool as
+orchestrator for the rest of the run and tries one other eligible pool
+(`decision.orchestrator_escalated`), and only then settles on a qualified
+`completed_with_concerns`/`blocked` outcome. Safety bounds (`maxActions`,
+`maxItemsPerExpansion`) still terminate immediately with a qualified outcome.
 `complete` still requires successful delivery and verification. `stop` returns
 `completed_with_concerns` with a ready best-effort artifact when useful work
 exists, or `blocked` when it does not; neither hides failed verification.

package/src/cli.js

@@ -17,7 +17,7 @@ import { release } from './lib/release.js';
 import { cmdWorkflow } from './workflow/cli.js';
 import { cmdStrategy, maybeRefreshStrategy } from './strategy-cli.js';
 import { cmdIntegrate, installIntegration } from './integrate.js';
-import { helpForArgs } from './help.js';
+import { helpForArgs, usageLine } from './help.js';
 import { resolveDispatchModel } from './lib/strategy.js';
 
 export function getBullswarmDir() {
@@ -96,6 +96,34 @@ async function cmdRun(opts) {
   }
   const targetDir = resolve(opts['add-dir'] ?? process.cwd());
 
+  // Validate task input before routing so malformed invocations never fall
+  // through to the caller-session path.
+  if (opts['task-file'] === true || opts.prompt === true ||
+      (opts['task-file'] != null && typeof opts['task-file'] !== 'string') ||
+      (opts.prompt != null && typeof opts.prompt !== 'string')) {
+    console.error('usage: --prompt and --task-file require a value');
+    return 2;
+  }
+  if (opts['task-file'] && opts.prompt != null) {
+    console.error('usage: choose one of --prompt, --task-file, or trailing task text');
+    return 2;
+  }
+  if (opts['task-file'] && opts.rest.length) {
+    console.error('usage: choose one of --task-file or trailing task text');
+    return 2;
+  }
+  if (opts.prompt != null && opts.rest.length) {
+    console.error('usage: choose one of --prompt or trailing task text');
+    return 2;
+  }
+  const taskText = opts['task-file']
+    ? readFileSync(opts['task-file'], 'utf8')
+    : opts.prompt ?? opts.rest.join(' ');
+  if (!taskText.trim()) {
+    console.error('empty task: pass --task-file, --prompt, or the task as arguments');
+    return 2;
+  }
+
   // Recursion guard FIRST — core-owned, env handshake.
   let state = loadState(getBullswarmDir());
   try {
@@ -166,15 +194,6 @@ async function cmdRun(opts) {
     subscription: poolView.subscription ?? connector.subscription ?? null,
   };
 
-  // Task text: --task-file content or stdin string.
-  const taskText = opts['task-file']
-    ? readFileSync(opts['task-file'], 'utf8')
-    : opts.rest.join(' ');
-  if (!taskText.trim()) {
-    console.error('empty task: pass --task-file or the task as arguments');
-    return 2;
-  }
-
   const stamp = `${Date.now()}-${Math.random().toString(36).slice(2, 7)}`;
   const runDir = join(getBullswarmDir(), 'runs');
   mkdirSync(runDir, { recursive: true });
@@ -455,9 +474,7 @@ export async function main(argv) {
     case 'release':
       return cmdRelease(opts);
     default:
-      console.error(
-        `unknown verb "${verb}". try: setup | integrate | run | health | pools | strategy | doctor | workflow | runs | version | release`,
-      );
+      console.error(`unknown verb "${verb}". Run "bullswarm --help" for the list of commands.`);
       return 2;
   }
 }
@@ -467,7 +484,7 @@ export async function main(argv) {
 function cmdRelease(opts) {
   const kind = opts.rest[0];
   if (!['patch', 'minor', 'major'].includes(kind)) {
-    console.error('usage: bullswarm release patch|minor|major [--dry-run]');
+    console.error(`usage: ${usageLine(['release'])}`);
     return 2;
   }
   try {