Agent Persona Exploration - 2026-07-15

[github-actions]

Persona Overview

• Agent: agentic-workflows
• Personas This Run: Program Manager, Designer, Legal / Compliance
• Scenarios Tested: 4
• Average Quality Score: 4.6/5.0

Key Findings

• The guidance strongly favors schedule + workflow_dispatch for non-technical digests and pull_request with paths: filters for governance-style PR checks.
• github via gh-proxy is the dominant read mechanism, while writes are consistently routed through add-comment or create-issue safe outputs.
• Security guidance is consistently strong: read-only agent jobs, minimal permissions, explicit noop behavior, and narrow scope filters.
• Non-technical personas are generally handled well when the task maps to recurring digests or narrow PR governance checks.
• Weaker spots appear in compliance scenarios where evidence sources and fallback heuristics are not spelled out enough.

Top Patterns

1. Most common triggers: schedule + workflow_dispatch, then pull_request with scoped paths:.
2. Most recommended tools: github (gh-proxy) plus safe outputs; specialized tools are rarely suggested unless clearly required.
3. Security practices observed: read-only permissions, scoped triggers, safe outputs for all writes, and explicit noop criteria.

View High Quality Responses (Top 2-3)

• Program Manager — weekly milestone digest: strong report-window, grouping, blocker-highlighting, and dedup guidance.
• Designer — Figma-link PR check: clear use of pull_request, file scoping, PR metadata reads, and add-comment output.

View Areas for Improvement (Top 2-3)

• Compliance license scan: good scoping, but SPDX evidence/fallback guidance is still fairly generic.
• Policy/disclosure digest: approval-evidence matching is not concrete enough for ambiguous cases.
• Design governance prompting: accepted design-reference formats could be specified more clearly.

Recommendations

1. Strengthen .github/aw/create-agentic-workflow-trigger-details.md with concrete heuristics for dependency/license evidence sources and fallback behavior.
2. Add an approval-evidence matching example for policy and SECURITY-file digests in .github/aw/github-agentic-workflows.md or .github/aw/workflow-patterns.md.
3. Document accepted design-reference patterns more explicitly in .github/aw/create-agentic-workflow.md to improve PR governance prompts.

Generated by 🎭 Agent Persona Explorer · 7.92 AIC · ⌖ 10.1 AIC · ⊞ 6.5K · ◷