Investigate: Evaluator pipeline — sequential vs parallel agent pattern

[dean0x]

Context

Extracted from #170 (v2.0.0 release) — deferred to post-v2.

The quality gate pipeline currently runs a sequential chain: Validator → Simplifier → Scrutinizer → re-Validate → Evaluator → Tester. The code review pipeline uses a different pattern: multiple Reviewer agents in parallel with different focus skills, then synthesis.

Question

Does the evaluation pipeline benefit from the same multi-agent parallel pattern? Or is the sequential chain correct because phases depend on prior outputs (e.g., Simplifier must run before Scrutinizer sees the code)?

Tasks

• Analyze which evaluation phases are truly sequential (output depends on prior phase) vs. parallelizable (independent quality checks)
• Determine which phases map to distinct focus skills (like reviewer focuses: security, architecture, etc.)
• Design the Evaluator agent with parameterized focus if the pattern fits
• Implement, or document why the current sequential approach is better for this pipeline
• Update implementation-orchestration skill and /implement commands if adopting