Post-flight validation for Output SDK workflow operations. Systematic verification of step completion, convention compliance, quality validation, and deliverable verification.
SkillScore96Maintained
Automatic
◆AI Engineer pack
Last verified 8/27/2026
What it changes
Passed 22 of 23 test cases with the skill, and 11 of 23 without it.
Without skill48%
With skill96%
Checks it passed that the model alone missed
prohibits let declarations in generated code
flags use of let in favor of const declarations
specifies providing CLI commands for the Output SDK
and 14 more in the full report
11 fixed11 passed either way1 still failing0 broke
Pass-rate lift
+48 pp
Tokens
+19%
Agent turns
0%
Response time
-34%
Model gemini-3.6-flashMeasured 8/27/2026v222 of 23 cases passed with skill
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 11 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 11
Prompt
Looking at function calls in our activities, we see `processData(payload)`. According to our codebase's formatting style, how should parentheses spaces be formatted for arguments?
What was checked, without and with the skill
Checked
Without skill
With skill
requires adding space inside function call parentheses like fn( x )
Missed
Passed
case-11 · gemini-3.6-flash · 8/27/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md104 linessingle file
Post-Flight Rules for Output SDK Workflows
Execution Verification
After completing all steps in the process_flow, systematically verify:
Step Completion Audit
Every numbered step has been read, executed, and delivered according to its instructions
All steps that specified a subagent were delegated to the correct subagent
If any subagent was not used as specified, document why and report to the user
If any step was not executed according to instructions, explain which part was misread or skipped
Output SDK Convention Compliance
Verify the following conventions were followed:
Import Conventions
All TypeScript/JavaScript imports use .js extension for ES modules
No direct axios usage - HttpClient wrapper used throughout
Proper import paths for Output SDK packages (@outputai/core, @outputai/llm, etc.)
Workflow Structure
Workflow exported in entrypoint.ts: export * from './path/to/workflow.js';
All external operations wrapped in Temporal activities (steps)
Proper error handling with ApplicationFailure patterns
Retry policies configured appropriately
Schema Placement
All schemas for aiSdk.Output.object() defined in types.ts, not inline in step functions
LLM output schemas use .describe() only -- no .min()/.max()/.length() on numbers or arrays
Prompt files do not contain JSON output format instructions when aiSdk.Output.object() is used
LLM Provider & Variables
All prompt files use the same provider (no mixing unless explicitly requested)
Structured generateText/Agent variables use Liquid loops, conditions, or dot notation appropriately
Code Style (see output-dev-code-style)
No trailing commas in any generated code
No let declarations -- all variables use const
Arrow functions use parens only when needed (multi-param or destructured)