Fleet-wide parallel device operations - concurrent health checks, config audits, routing snapshots, severity-sorted reporting, and failure-isolated multi-device automation.
SkillScore68Maintained
Automatic
◆DevOps & SRE pack
Last verified 8/16/2026
What it changes
Passed 14 of 20 test cases with the skill, and 7 of 20 without it.
Without skill36%
With skill68%
Checks it passed that the model alone missed
Marks R4 as CRITICAL due to missing or zero routes.
Executes all 4 device checks concurrently in a single parallel wave.
Issues multiple pyats_run_show_command calls simultaneously in one response.
and 9 more in the full report
7 fixed8 passed either way7 still failing0 broke
Pass-rate lift
+32 pp
Tokens
+122%
Agent turns
0%
Response time
-39%
Model gemini-3.6-flashMeasured 8/16/2026v115 of 22 cases passed with skill · 20 comparable
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 6 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 6
Prompt
We collected route tables from R1, R2, R3, and R4. R1 has 47 total routes (5 connected, 12 OSPF, 28 BGP), R2 has 45 total, R3 has 38 total, and R4 returned 0 routes. Format this into a fleet routing summary table.
What was checked, without and with the skill
Checked
Without skill
With skill
Displays a summary table with columns for Device, Total, Conn, OSPF, BGP, Default Rte, and Status.
Missed
Passed
Marks R4 as CRITICAL due to missing or zero routes.
Missed
Passed
case-11 · gemini-3.6-flash · 8/16/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md311 linessingle file
Parallel Fleet Operations
When to Use
Fleet-wide health checks across all devices in the testbed
Mass configuration audits (collect running configs from every device)
Network-wide routing table snapshots for baseline or comparison
Pre/post change validation across all affected devices simultaneously
Any operation where running sequentially on 10+ devices would be too slow
How pCall Works in OpenClaw
In OpenClaw, parallel execution (pCall) is achieved by listing multiple exec commands in a single response. The agent runtime dispatches them concurrently and collects all results before proceeding.
Key Principles
Group by role or site -- Organize devices into logical groups (core, distribution, access, WAN, DC) before dispatching
Run operations concurrently -- List one MCP call per device; they execute in parallel
Failure isolation -- If one device times out or errors, the other results are still collected
Result aggregation -- Collect all results, then produce a unified fleet report
Severity sorting -- Sort findings from CRITICAL to HEALTHY so the worst problems surface first
pCall Pattern
To run the same command on multiple devices in parallel, list the calls together: