mcpbeat

Reader Protocol Pr Review

automattic/reader-protocol-pr-review

Use when reviewing PRs that touch the Reader's multi-protocol social surfaces (`client/reader/social/`, `client/reader/atmosphere/`, `client/reader/mastodon/`, `packages/api-core/src/reader-{atmosphere,mastodon}/`, `packages/api-queries/src/reader-{atmosphere,mastodon}.ts`). Captures the review rubric + the recurring smells from the CM-625 / CM-658 / CM-660 / CM-662 slice cycle.

4k tokens
context cost
the whole folder, loaded on every use
2
files
instructions only
0
copies elsewhere
how many repositories repackaged it
12634
stars on the repo
on the repository, not the skill itself

Install

one command, takes just this skill from the repository
npx skills add https://github.com/Automattic/wp-calypso --skill reader-protocol-pr-review

What comes with it

9 875 bytes besides the instruction
references/common-smells.md

The instruction itself

8 sections, as written by the author

Reader Protocol PR Review

Use this when reviewing a PR in the Reader's social/atmosphere/mastodon territory — slice work for the protocol-adapter pattern, optimistic mutations, or the shared post-card components. The recurring findings are concrete enough to be a checklist; the rubric below comes from the CM-658/CM-660 cycle where the same issues appeared in both PRs.

Rubric

For each finding write:

| Severity | File:line | Issue (1-2 sentences) | Suggested fix |
  • Severity tags: blocker (PR can't merge), important (real bug, would cause user-visible issues), nit (cleanup or hardening).
  • File:line: anchor with absolute path or repo-relative path so the comment can be left inline on GitHub. Use gh api repos/Automattic/wp-calypso/pulls/{n}/comments to post inline replies (not top-level PR comments) when the line is in the diff.
  • In-PR vs follow-up: explicitly call this out. Architectural changes that span multiple slices belong in a follow-up issue. Defensive hardening, missing default arms, etc. are cheap to fold in.
  • Verdict line: Approve / Request changes / Needs more discussion. If approve, list any follow-ups worth filing as separate tickets.

Cap reviews at ~700 words. The reviewer's signal-to-noise ratio matters — detailed prose drowns out the actual blockers.

When to dispatch the code-reviewer agent

The superpowers:code-reviewer agent is independent and reads the diff fresh. Dispatch it when:

  • The PR is large (>20 files or >500 lines).
  • The PR shape is structurally similar to a sibling PR you've already reviewed (it'll surface the same recurring smells consistently).
  • You authored adjacent code and want an independent eye.

Brief the agent with: PR number, branch path, what the PR does, and a list of recurring smells from references/common-smells.md to specifically check. The agent's response is the body of your review.

Reply convention on PR threads

  • No "thanks for the review"superpowers:receiving-code-review rules apply. State the fix or push back; actions speak.
  • Reply on inline comment threads for inline review comments (gh api repos/{owner}/{repo}/pulls/{pr}/comments/{id}/replies), not as top-level PR comments. The threading matters for context.
  • Top-level comment for review-summary acknowledgement, listing which findings landed in commit X and which are deferred follow-ups (with rationale).
  • No fabricated wp.me shortlinks — never invent a shortlink for a *.wordpress.com URL. See a8c-url-shorthand:a8c-url-shorthand.

Common smells

See references/common-smells.md — checklist of 9+ recurring findings from the recent slice cycle. Walk it before submitting your review.

Cross-pollinating fixes between sibling PRs

When two slice PRs are open simultaneously (CM-658 + CM-660 was the canonical case), review feedback on one almost always applies to the other. The architectural patterns mirror each other.

After landing a fix on PR A, check whether PR B has the same issue:

git -C /Users/pfefferle/Code/wp-calypso show <fix-commit-on-A> -- <relevant-paths>
git -C /Users/pfefferle/Code/wp-calypso show <branch-of-B>:<relevant-path> | grep <smell>

If yes, apply the fix to B too. Don't wait for the reviewer to file the same finding twice.

File-shape expectations

These are the Reader-protocol files that consistently need attention. Reviewers should check each:

| File pattern | What to verify |

|---|---|

| client/reader/social/components/post-card/{like,repost}-context.tsx | LikeAction/RepostAction interface has no dead fields; null-action default mirrors the live shape. |

| client/reader/social/components/post-card/{like,repost}-button.tsx | Renders a static-count fallback when !action.supportednot null. The fallback lives inside the button, not in <PostCardCounts>. |

| client/reader/social/components/post-card/post-card-counts.tsx | Always renders <LikeButton> / <RepostButton> (no connectionId ternary for the like/repost slot — the button handles supported / unsupported internally). |

| client/reader/{atmosphere,mastodon}/use-{protocol}-{like,repost}-action.ts | errorMessageFor* switch has a default: arm; trackError includes a logToLogstash call; cid/rkey guards before calling mutate. |

| packages/api-queries/src/reader-{atmosphere,mastodon}.ts | Optimistic patcher is connection-scoped; cancelQueries is wrapped in try/catch in onMutate; mutation factories take connectionId: number, not QueryClient. |

| packages/api-core/src/reader-{atmosphere,mastodon}/fetchers.ts | Path-interpolated wire IDs are encodeURIComponent-wrapped. |

Reference

  • references/common-smells.md — the smell catalog, with concrete examples from the slice cycle.
  • client/reader/social/AGENTS.md — describes the provider-adapter pattern + naming convention.
  • client/reader/AGENTS.md § "Optimistic-mutation hardening checklist" — engineering-side pre-flight for the patterns this skill reviews.

How to use it

Copy the folder

Take automattic/reader-protocol-pr-review from the repository into ~/.claude/skills for personal use, or into .claude/skills inside a project.

Check the name does not clash

The agent identifies a skill by the name field in its header. Two skills with the same name cannot sit side by side — one of them will be ignored.