What problem does it solve?
It prevents expensive rework during coding by orchestrating targeted implementation-time review subagents and turning their findings into clear next actions.
Core Features & Use Cases
- Implementation-time, not PR-time review: focuses on next 1–5 moves for in-progress diffs and avoids creating PR comments.
- Structured context acquisition: centralizes what the main agent must collect (diff, changed symbols/files, phase, test status, logs, project rules, plus optional dynamic/historical/spec consistency context).
- Risk-flag driven agent routing: selects a small set of specialist reviewers (typically up to 4, up to 6 for high-risk changes) based on phase and risk flags, then aggregates results into blocking/important/next_actions.
- Finishing-phase validation guardrails: ensures validation gaps are surfaced and targets the meta-reviewer when needed for safe continuation.
- Accountability and deduped aggregation: uses the orchestration accountability format, consolidates duplicate findings, and identifies concentrated risks.
Quick Start
Ask your AI coding main agent to run implementation-review-orchestration for the current in-progress diff, providing current phase, changed files/symbols, test status, any error logs, and available project_rules so it can select appropriate implementation-time subagents and return internal next actions.
