Skip to content

PHASE5_RESEARCH_BACKLOG_2026_04_29.md

Latest commit

 

History

History
38 lines (25 loc) · 1.05 KB

PHASE5_RESEARCH_BACKLOG_2026_04_29.md

File metadata and controls

38 lines (25 loc) · 1.05 KB
Raw

Phase 5 Research Backlog: 2026-04-29

Scope

Phase 5 intentionally avoids large architecture changes. Each research issue gets a small proof plan and explicit evidence gate before implementation can change production review behavior.

Issue Plans

Use:

agora research plan
agora research plan --json

The command reports one experiment plan for:

  • #466 Binary internal severity
  • #469 Ambiguous-case calibration dataset
  • #471 Cross-file interaction review
  • #481 Bandit exploration beyond configured pool

Guardrails

  • No public severity model change without benchmark evidence.
  • Ambiguous fixtures stay separate from golden-bug recall fixtures.
  • No InteractionReviewer path until cross-file missed-bug fixtures prove value.
  • No live model swap from catalog exploration without passing the reference benchmark gate.

Verification

Required targeted checks:

  • pnpm exec vitest run packages/core/src/tests/research-experiments.test.ts
  • node packages/cli/dist/index.js research plan --json
  • pnpm typecheck