Agent Persona Exploration - 2026-09-04

### Persona Overview
- **Agent**: agentic-workflows
- **Personas This Run**: Program Manager, Designer, Legal / Compliance
- **Scenarios Tested**: 4
- **Average Quality Score**: N/A

### Key Findings
- Persona selection correctly prioritized non-technical roles that were still unexplored.
- Generated scenarios aligned well with documented gh-aw persona-to-trigger patterns for Program Manager, Designer, and Legal / Compliance use cases.
- Direct agent response testing was unavailable because the installed invocation surface for ad hoc evaluation was not present in this environment (`gh copilot` missing).
- Quality scoring is unavailable for tested scenarios because no agent-generated recommendations were returned.

### Top Patterns
1. Scheduled workflows fit recurring digests and compliance drift audits.
2. PR-scoped workflows with `paths:` fit design-token governance and dependency-license review.
3. Safe outputs should remain narrow: `create-issue` for recurring reports/escalations, `add-comment` for PR-local findings, and explicit `noop` when no action is required.

View High Quality Responses (Top 2-3)

- No high-quality response samples available this run because invocation failed before scenario evaluation.

View Areas for Improvement (Top 2-3)

- Provide a documented local or CLI-accessible ad hoc evaluation path for the `agentic-workflows` custom agent when chat-mode invocation is unavailable.
- Clarify fallback guidance in `.github/aw/create-agentic-workflow.md` for research workflows that need to assess agent behavior without creating files.
- Add an example in `.github/aw/github-agentic-workflows.md` showing how to record `invocation_unavailable` results consistently in persona research runs.

### Recommendations
1. Add a documented evaluation fallback to `.github/aw/create-agentic-workflow.md` for environments where conversation-mode invocation is unavailable.
2. Expand `.github/aw/github-agentic-workflows.md` with a concise persona-research example covering partial-result handling and `invocation_unavailable` classification.
3. Surface a stronger note in `.github/aw/safe-outputs.md` or related guidance that research workflows should still publish partial findings even when agent testing cannot run.

View Scenario Inventory and Invocation Result

- **Program Manager** — Responsibilities: Coordinates cross-team milestones, schedules, and dependency tracking; Pain points: Cross-team reporting and blocked-item tracking are time-consuming
- **Designer** — Responsibilities: Manages design systems, accessibility checks, visual asset reviews; Pain points: Token governance and design-link validation are inconsistent
- **Legal / Compliance** — Responsibilities: Tracks license compliance, policy files, and security disclosures; Pain points: Dependency/policy review is tedious and needs consistent escalation rules

| ID | Persona | Task | Expected Workflow Type | Status |
|---|---|---|---|---|
| PM1 | Program Manager | Weekly cross-team milestone status digest with blocked items highlighted | Scheduled | invocation_unavailable |
| PM2 | Program Manager | Escalate missed dependency due dates from active projects into a rolling coordination issue | Scheduled | generated_only |
| DS1 | Designer | Flag PRs that modify shared design tokens or component CSS without a matching Figma or spec link | PR automation | invocation_unavailable |
| DS2 | Designer | Review PRs for accessibility-risk UI changes and summarize likely issues | PR automation | generated_only |
| LC1 | Legal / Compliance | Scan new dependencies added in PRs for non-permissive SPDX licenses | PR automation | invocation_unavailable |
| LC2 | Legal / Compliance | Run a monthly policy drift audit for required compliance control files | Scheduled | invocation_unavailable |

- **Invocation error**: `! Copilot CLI not installed`

> Generated by [🎭 Agent Persona Explorer](https://github.com/github/gh-aw/actions/runs/33889879825) · pi · gpt54 · 28 AIC · ⊞ 7.9K · [◷](https://github.com/search?q=repo%3Agithub%2Fgh-aw+is%3Aissue+%22gh-aw-workflow-call-id%3A+github%2Fgh-aw%2Fagent-persona-explorer%22&type=issues)

Read the recommendations against .github/aw/create-agentic-workflow.md, .github/aw/github-agentic-workflows.md, and .github/aw/safe-outputs.md. Document an evaluation fallback, add the persona-research partial-result example, and clarify that research workflows should publish partial findings when invocation is unavailable. Done means all three requested guidance updates are covered in the named files.

Written by the indexing model from the issue text.