Runs two parallel agents — one performs a rigorous PR review (via the review-pr skill), the other evaluates findings against the project's design context (ADRs, sprint goals) — then synthesizes a decision-ready report with Apply / Skip / Discuss verdicts per finding. Copied from zora-pantheon `.claude/commands/review-and-evaluate.md` and adapted to this repo's skill structure.
81%
Does it follow best practices?
Run evals on this skill
Adds up to 20 points to the overall score
Low
Low-risk findings worth noting
Runs two parallel agents — one performs a rigorous PR review, the other evaluates findings against the project context — then synthesizes a decision-ready report.
git fetch origin to ensure remote refs are up to date. Do NOT rebase or modify the current branch.git branch --show-current to identify the current branch.git merge-base origin/main HEAD to find the common ancestor.git log --oneline <merge-base>..HEAD to list ONLY the commits unique to this branch.git diff <merge-base>..HEAD --name-only to get the list of changed files.Launch a general-purpose agent with the following task:
Read the review guidelines from
.agents/skills/review-pr/SKILL.mdand follow them exactly, with these overrides:
- Branch scope: Use
<merge-base>from step 0 above. Only review files changed between<merge-base>andHEAD.- Do NOT post anything to GitHub. Skip steps 6 and 7 entirely. Just produce the severity-based summary report (step 5).
- Do NOT ask which items to post. The synthesis step below handles that.
Launch a general-purpose agent with the following task:
You are a sprint / project alignment analyst. Your job is to understand the goals and design decisions behind the current branch's work, then produce an analysis that will be used to evaluate PR review findings.
Steps:
- Read the commit messages from
<merge-base>..HEADto understand what was built and why.- Read the conversation context provided by the user (if any) to understand the broader project goals.
- Read the actual implementation files (changed files in branch scope).
- Read the existing pattern files that the new code is supposed to follow (identify these from imports, directory siblings, and
CLAUDE.mdreferences).- Read
docs/adr/(standalone ADR files),docs/spec.html(foundational ADR-001 through ADR-030), anddocs/diary.md(dated entries for decisions not yet promoted to standalone ADRs) to understand decisions that constrain the design space (e.g., ADR-013 security headers, ADR-014 SW block, ADR-024 no fp-ts, the Conventional Commits + rebase-merge convention, ADR-0035 bot-merge workflow).- Produce a detailed analysis:
- What was built: Summarize each logical unit of work from the commits.
- Pattern compliance: For each new file, compare against the existing pattern it mirrors. Note matches and deviations.
- Justified deviations: Identify deviations that are improvements over the existing pattern (e.g., better error handling, stronger typing).
- Unjustified deviations: Identify deviations that seem accidental or that break consistency without clear benefit.
- Cross-cutting concerns: Issues that span multiple files or features.
- Risk assessment: Things that could cause problems downstream.
After both agents complete, combine their outputs into a single decision-ready report with this structure:
(From Agent 1)
For EACH finding from Agent 1, present:
| Item | Action | Effort | Reasoning |
|---|---|---|---|
| ... | Apply / Skip | ~X min | One-line explanation |
End with: "Which items would you like me to apply?"
(Note: when invoked from /pr-iterate, the calling skill consumes the verdicts directly — Apply items get applied automatically; Discuss items become the iteration's escalation list; Skip items are recorded and ignored. The interactive question is bypassed in that mode.)
agranado2k/ai-report-platform
caeb0da
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.