chore: align AI agent roster with EFT Viewer scope

[wct097]

Summary

• Trim .claude/agency-config.json to the six categories relevant to a C#/.NET desktop fingerprint viewer: design, engineering, product, project-management, specialized, testing.
• Remove 11 non-aligned category directories wholesale (academic, examples, finance, game-development, integrations, marketing, paid-media, sales, spatial-computing, strategy, support) and prune individual off-scope profiles inside the kept categories (web frontend, blockchain, embedded firmware, Feishu/WeChat, API testing, brand/whimsy design, Jira/studio PM, etc.).
• Expand .claude/agent-teams.json from 2 to 26 shortcodes covering the roles this project actually uses: desktop/parser engineering, mobile (future MAUI), security, DevOps/releases, technical writing, git workflow, code review, onboarding, minimal-change edits, rapid prototyping, reality checks, performance (WSQ decode), accessibility, test analysis, evidence collection, Avalonia UI/UX, inclusive visuals, product/sprint management, and document generation.

Rationale

Initial agency-agents sync pulled 179 personas across 17 categories. Most were irrelevant to an OSS Avalonia desktop app parsing ANSI/NIST-ITL biometric files, and kept noise out of the agent registry. This PR narrows the roster to a curated set that maps cleanly to this project's surface area.

Test plan

• Confirm .claude/agents/ contains only aligned categories and profiles
• Confirm no empty directories under .claude/agents/
• Confirm .claude/agent-teams.json shortcodes resolve to existing files
• Confirm .claude/agency-config.json lists only the six retained categories

🤖 Generated with Claude Code

[wct097]

Code review — approve with minor suggestions

Config-only scope-reduction PR. Every checklist item passes on HEAD.

Fact-check

• Both JSON files parse cleanly, end with trailing newlines.
• synced_categories matches dirs on disk (design, engineering, product, project-management, specialized, testing).
• All 26 shortcode paths resolve to existing files.
• No empty directories under .claude/agents/.
• 28 personas on disk (matches PR claim).
• No secrets committed.

Suggestions

1. Two personas are orphaned (on disk but not registered in agent-teams.json):

   • .claude/agents/testing/testing-tool-evaluator.md
   • .claude/agents/testing/testing-workflow-optimizer.md

   Register or prune.

2. MOB scope question — retained for the future .NET MAUI roadmap per CLAUDE.md; defensible, just flag in description next time.

3. Optional: no dedicated data/binary-format persona for ANSI/NIST-ITL + WSQ work. SE+SA cover it for a solo dev; flag for future consideration.

Nits

• Key grouping in agent-teams.json is by category (not alpha); grouping reads better — preserve on future edits.

Recommendation: address suggestion #1, then merge.

[wct097]

Pushed fix for suggestion #1: registered TE (tool-evaluator) and WO (workflow-optimizer) so every on-disk persona resolves through the team registry. Suggestions #2 (MOB retained for MAUI roadmap) and #3 (parser specialist) deferred per solo-dev context.