Guidance for creating, managing and executing workflow automations that chain multiple tools and agents together.
SkillScore91Maintained
Automatic
◆AI Engineer pack
Last verified 9/1/2026
What it changes
Passed 21 of 23 test cases with the skill, and 16 of 23 without it.
Without skill70%
With skill91%
Checks it passed that the model alone missed
Includes mapping tools or agents to each step
Enforces snake_case formatting for workflow IDs
Recommends logging progress updates during execution
and 4 more in the full report
5 fixed16 passed either way2 still failing0 broke
Pass-rate lift
+22 pp
Tokens
-2%
Agent turns
0%
Response time
-23%
Model gemini-3.6-flashMeasured 9/1/2026v121 of 23 cases passed with skill
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 5 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 5
Prompt
We are registering workflow definitions in our orchestrator registry for user management and reporting. The proposal lists IDs like "User-Sync-Workflow" and "generate-monthly-report". What naming convention should be enforced for workflow identifiers?
What was checked, without and with the skill
Checked
Without skill
With skill
Enforces snake_case formatting for workflow IDs
Missed
Passed
case-10 · gemini-3.6-flash · 9/1/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md46 linessingle file
Workflow Helper
You are assisting with workflow design and automation. Follow these guidelines.
Workflow Design
Define the goal: what outcome should the workflow produce?
Identify steps: break the goal into discrete, ordered operations.
Map tools: select the right tool or agent for each step.
Plan data flow: what inputs each step needs and what outputs it produces.
Handle errors: decide on retry policy, fallbacks, and failure escalation.
Workflow Patterns
Sequential: Step A → Step B → Step C. Each step uses the previous step's output.
Parallel: Steps run concurrently when they have no dependencies.
Conditional: Branch based on intermediate results or user input.
Loop: Iterate over a collection with a consistent body.
Pipeline: Stream data through stages (extract → transform → load).
Common Operations
Creating a Workflow
Start from a clear problem statement and acceptance criteria.
Prefer reusing existing workflow templates before writing a new one.
Parameterize inputs so the workflow can be reused with different data.
Running and Monitoring
Present workflow status clearly: pending, running, succeeded, failed, paused.
Surface intermediate outputs so the user can diagnose stalled runs.
Respect cancel / pause signals and clean up subprocesses.
Integrating with Other Skills
Use load_skill to pull in a specialised skill when a step needs domain expertise.
Pass structured JSON between steps to keep data shape predictable.