rfe.review¶
Score and improve RFEs with a multi-phase agent pipeline. Accepts one or more Jira keys (RHAIRFE-NNNN) or local IDs (RFE-NNN); missing RFEs are fetched from Jira first. The orchestrator never reads RFE content directly -- all content-heavy work is delegated to background sub-agents (fetch, assess, feasibility, review, revise) launched in parallel waves and polled via scripts/check_review_progress.py. It runs rubric-based assessment (assess-rfe / rfe-scorer subagent), launches per-RFE feasibility checks (rfe-feasibility-review), synthesizes scored review files, auto-revises failing RFEs (filter_for_revision.py), and re-assesses up to 2 cycles, preserving cumulative scores and revision history across cycles. Can return headlessly to a calling skill (auto-fix or split) or print an interactive summary with next-step suggestions.
Plugin: rfe-creator | User-invocable
Contract¶
Score RFEs against the rubric, check technical feasibility, and auto-revise failing RFEs across up to two re-assessment cycles.
Identity
- Produces a review file with rubric scores, a feasibility verdict, and a recommendation per RFE.
- Auto-revises failing RFEs and re-assesses up to 2 cycles.
Optimization Targets
Invariants
- Never read RFE bodies into orchestrator context — delegate to agents and read only frontmatter.
- Do not exceed 2 re-assessment cycles.
Traceability
Diagram¶
Arguments¶
/rfe.review <ID> [ID2 ...] [--headless] [--caller <name>]
| Argument | Required | Default | Description |
|---|---|---|---|
ID |
- | One or more space-separated RFE IDs (RHAIRFE-NNNN or RFE-NNN) | |
--headless |
- | Suppress end-of-run summary; used when called from rfe.auto-fix or rfe.split | |
--caller |
none |
Identifies calling skill for headless return routing |
Usage¶
/rfe.review RHAIRFE-1234
/rfe.review RFE-001 RFE-002 RFE-003
/rfe.review --headless --caller autofix RHAIRFE-1234 RHAIRFE-5678