Objective

Codify the workflow_run vs pull_request trigger selection reasoning as a reusable decision tree pattern in the create-agentic-workflow.md prompt.

Context

From discussion #21704 (Agent Persona Exploration - 2026-03-19), the evaluation found that the agent correctly chooses between workflow_run and pull_request triggers — but this reasoning is implicit. Making it explicit in the prompt library will help new workflow authors understand the tradeoff.

Key insight observed during evaluation:

• Use workflow_run when CI artifacts are needed (coverage reports, build outputs, deployment results)
• Use pull_request with path filters when reacting to code changes directly (schema reviews, visual regression)

Approach

1. Locate the create-agentic-workflow.md prompt file (likely in skills/ or .github/)
2. Add a "Trigger Selection" section with a decision tree:
   • Does the workflow need CI artifacts from a prior run? → workflow_run
   • Is the workflow reacting to code changes (files, PRs)? → pull_request with paths: filters
   • Is it time-based or manual? → schedule + workflow_dispatch
   • Is it release/tag-based? → push: tags:
3. Include the trigger hierarchy pattern: workflow_run → pull_request (path filters) → schedule + workflow_dispatch → push: tags
4. Add a brief rationale for each choice with an example snippet

Files to Modify

• The create-agentic-workflow.md prompt (check skills/ and .github/ directories)

Acceptance Criteria

• Decision tree for trigger selection is documented in the prompt
• Includes rationale for workflow_run vs pull_request choice
• Covers all 4 trigger categories with examples
• New workflow authors can use it without external guidance

Generated by Plan Command for issue #discussion #21704 · ◷

• expires on Mar 21, 2026, 1:48 AM UTC