deploy-multi-agent-swarm — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited deploy-multi-agent-swarm (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
| Aspect | Play 07 (Supervisor) | Play 22 (Swarm) |
|---|---|---|
| Coordination | Central supervisor decides | Agents bid and vote |
| Task assignment | Top-down delegation | Competitive bidding |
| Quality control | Supervisor aggregates | Consensus voting |
| Failure handling | Supervisor reassigns | Swarm self-heals |
| Scalability | Limited by supervisor | Horizontally scalable |
| Single point of failure | Supervisor is SPOF | No SPOF |
az account showaz bicep lint -f infra/main.bicep
az deployment group create -g $RG -f infra/main.bicep -p infra/parameters.json{
"swarm": {
"agents": [
{ "role": "researcher", "model": "gpt-4o-mini", "instances": 3 },
{ "role": "analyst", "model": "gpt-4o", "instances": 2 },
{ "role": "writer", "model": "gpt-4o-mini", "instances": 2 },
{ "role": "validator", "model": "gpt-4o-mini", "instances": 2 }
],
"bidding": { "timeout_ms": 5000, "min_bids": 2 },
"consensus": { "method": "majority_vote", "min_votes": 3, "quorum": 0.6 },
"max_parallel": 5,
"max_rounds": 3
}
}| Phase | Duration | What Happens |
|---|---|---|
| Task broadcast | 0s | Task published to all agents via Service Bus |
| Bidding window | 0-5s | Agents assess task, submit bid with confidence score |
| Selection | 5s | Top N bids selected (by confidence × past accuracy) |
| Execution | 5-30s | Selected agents execute in parallel |
| Voting | 30-35s | All agents vote on best result |
| Consensus | 35s | Majority-approved result returned |
| Method | Min Agents | Use Case |
|---|---|---|
| Majority vote | 3+ | General-purpose decisions |
| Weighted vote | 3+ | Expert agents get more weight |
| Unanimous | All | Safety-critical decisions |
| First-past-threshold | 1+ | Speed-optimized (accept first good-enough) |
swarm:{task_id}:bids, swarm:{task_id}:votes| Issue | Cause | Fix |
|---|---|---|
| No bids received | Agents not subscribed to topic | Check Service Bus subscriptions |
| Always same agent wins | Bidding not diverse | Rotate winners, add cooldown |
| Consensus never reached | Quorum too high (1.0) | Lower to 0.6 |
| Conflicting results | No conflict resolution | Add validator vote as tiebreaker |
| High cost | Too many agents bidding | Reduce swarm size, use fewer roles |
| Slow response | Sequential voting | Parallelize bid + vote phases |
Task → Service Bus (broadcast) → All Agents (bid) → Selection → Parallel Execution
↓
Results → Consensus Vote → Best Output
↓
Disagreement? → Extra Round or Human Escalation~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.