Reduce Import Workflow AI credit usage by setting engine.model: small

[Copilot]

The Import Workflow from Prompt Repository automation exceeded its max-ai-credits guardrail (1K) during a production run (~1.4K AIC). This change lowers model cost for that workflow so normal imports are less likely to breach the budget limit.

• Workflow frontmatter change

  • Updated .github/workflows/import-workflow.md to explicitly use the small model:

  engine:
    model: small

• Compiled workflow update

  • Regenerated .github/workflows/import-workflow.lock.yml so runtime config reflects the new model selection in agent/detection execution paths and metadata.

• Scope

  • No behavioral prompt rewrite or safe-output policy changes; this PR is limited to model-cost control for the import workflow.

[pelikhan]

@copilot use inlined agents to summarize the imported workflows and process them separately

[Copilot]

@copilot use inlined agents to summarize the imported workflows and process them separately

Implemented inlined-agent orchestration in the import workflow: it now delegates summarization to a dedicated inlined agent and processes workflow/custom-agent adaptations in separate inlined agent calls, with explicit consistency checks before writing files. Addressed in commit c00e537.