Evaluate alternatives to Copilot for blizz triage/implementation agents

[jerrod-storm]

Context

The org-level blizz workflows (blizz-triage.yaml, blizz-implement.yaml) currently use a Copilot-based agent for issue triage and PR implementation. These run on GitHub's infrastructure regardless of runner choice, meaning runner improvements (e.g. SHSH-28) don't help with Copilot's reliability or performance.

Proposal

Evaluate replacing the Copilot-based agent with an alternative:

• Cursor agents -- tight IDE integration, already in our workflow
• Linear's AI -- native to our issue tracker, could handle triage natively
• Anthropic's coding agent -- Claude-based, could be self-hosted or API-driven

Scope

1. Audit what the blizz workflows actually do today (triage logic, implementation triggers, PAT usage)
2. Evaluate alternatives on reliability, cost, and integration friction
3. Prototype a replacement for one of the two workflows
4. Migrate both workflows if the prototype works

Files affected

• .github/.github/workflows/blizz-triage.yaml
• .github/.github/workflows/blizz-implement.yaml
• Any consumer repos that pass secrets to these workflows