Compare RPI planning and implementation evidence and record severity-graded review findings.
npx skills add https://github.com/microsoft/hve-core --skill rpi-review
Please help me install this Agent Skill. Skill: rpi-review Source: https://github.com/microsoft/hve-core/tree/main/.github/skills/rpi/rpi-review Command: npx skills add https://github.com/microsoft/hve-core --skill rpi-review
After an implementation finishes, teams need an evidence-based acceptance review that separates whether work ran from whether the result is acceptable, and that routes every gap to the right follow-up stage instead of losing findings in chat.
.copilot-tracking/reviews/logs/ comparing plans, phase details, critiques, and changes records against acceptance criteria.RV-xxx findings to distinct destinations: defects to rpi-implement, decision gaps to rpi-plan, evidence gaps to rpi-research, and residual work to follow-up items.Ask the agent to run an RPI review of the completed task using its plan, phase details, critique, and changes record.
High-intent search queries and answers about installing and using this skill.
Invoke the rpi-review skill with the task's plan, phase details, plan critique, and changes record paths. It reconciles all artifacts, compares evidence against acceptance criteria, and writes one review record with severity-graded findings.
Execution status (Complete, Partial, Blocked) records whether planned work ran, while outcome (Conformant, Defects found, Residual work, Not accepted) records whether the result is acceptable. The review keeps these vocabularies strictly separate.
The review record is written to `.copilot-tracking/reviews/logs/{{YYYY-MM-DD}}/{{task_slug}}-review.md` using the bundled review-log template. Only this one canonical record may be created or updated during review.
No. The review stage is read-only for all artifacts except its own review record. Defects are routed to a later rpi-implement invocation, decision gaps to rpi-plan, and evidence gaps to rpi-research.
It stops as Blocked when a reviewable artifact set cannot be formed or evidence is insufficient for a credible verdict. It stops as Not accepted when material defects or unaccepted decision gaps remain after comparison.
abhiroopbGenerate phased Markdown implementation plans with code snippets and success criteria.
1molchuanExecute RPI implementation plans with parallel task routing and state updates.
EveryIncReviews requirements and plan documents with role-specific reviewer personas and synthesized findings.
BRANDNEWSHVTAutomate evidence-based code review for pull requests and feature branches.