parallel-review
Skill4 starsChanged 5 days agoMITView on GitHub
Catch architecture-compliance and line-level issues on a feature branch before the PR is opened, by running Gates' Gate 10 checks, Review's design/code-quality checklists, and a code-review pass against the branch diff ahead of time. Invoke manually after a feature completes and before gates.
4 other projects use a version of this.
Adopted by
Most recently taken up by
- danieleteti/delphimvcframeworktheir own version, first seen Nov 2, 2026
- stoa-platform/stoatheir own version, first seen Nov 1, 2026
- kostiantyn-matsebora/canopy-examplestheir own version, first seen Oct 11, 2026
- bhagyamudgal/skillstheir own version, first seen Sep 9, 2026
Other versions4 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 4
- danieleteti/delphimvcframework★ 1.4ksame name
- stoa-platform/stoa★ 9same name
- bhagyamudgal/skills★ 2same name
- kostiantyn-matsebora/canopy-examples★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository16 other files it gives its agents.
- benchmarkSkill
- bugfixSkill
- designSkill
- deterministic-pr-gatesSkill
- featureSkill
- gatesSkill
- pipeline-reviewSkill
- planSkill
- pr-followupSkill
- releaseSkill
- reviewSkill
- specSkill
- statusSkill
- sync-workflowSkill
- testSkill
- trim-contextSkill
The file
We split this skill into the rules it gives an agent
3 rules. 0 are common, 1 is distinctive, and 2 are very specific to this project.
0
Common
Most of the 3 edited versions we read keep these.
1
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 18
Read AGENTS.md/CLAUDE.md first — it defines the architecture rules enforced below.
- Very specific: line 37
Run the code-review:code-review skill against git diff develop...HEAD.
- Distinctive: line 63
Check 1 and Check 2 have both reported (or Check 2's fallback warning was printed instead), one of the three Combined-verdict strings above …
“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Parallel Review Agent
You are the Parallel Review Agent for an iOS app project. Your job is to catch architecture-compliance and line-level issues on a feature branch before the PR is opened, by running /gates' Gate 10 checks, /review's Design/Code-quality checklists, and code-review:code-review against the branch diff ahead of time.
Exception: Check 2 below follows /pr-followup's invocation-error fallback (don't stall, print a warning, continue), but prints its own warning string: it runs the code-review:code-review plugin, while /pr-followup runs the built-in code-review, so the two strings name different tools. Check 2's string is the only copy of it.
Trigger
Invoked manually after /feature completes and before /gates (e.g. /parallel-review feature/recurring-transactions). Defaults to the current branch if no argument is given.
Process
Read AGENTS.md/CLAUDE.md first — it defines the architecture rules enforced below.
Also read the following files if they exist — skip silently if absent:
.claude/context/invariants.md— project invariants; these supplement AGENTS.md/CLAUDE.md rules.claude/context/rejections.md— past violations on this project; flag any repeats as HIGH severity, except a repeat of an entry that recorded only a wording or style issue, which keeps its own severity. Rate only against entries from earlier PRs: if this branch already has an open PR, skip that PR's own entries (/review's isolated reviewer likewise sees only earlier PRs' entries)
Check 1 — Architecture compliance (/gates' Gate 10, pre-gates mode)
/review's own Architecture section defers to /gates having already run and expects a PR gate summary to check against — neither exists yet at this pre-PR, pre-/gates point, so run the actual checks instead of that deferral:
- Gate 10 — Architecture & layer-rule compliance:
scripts/check_gates.sh --only 10, with Gate 10's pass/fail criteria from.claude/skills/gates/SKILL.md. Its layer-rule checks read the committedgit diff develop...HEAD, the same scope this command needs, but its UI-selector check scans every tracked UI-test and app file (git ls-files), so a selector failure can come from code this branch didn't touch;/gatesfails on it all the same. It stops with exit 3 on a dirty tree, so commit first. - Design compliance checks, from
.claude/skills/review/SKILL.md, scoped togit diff develop...HEAD— only ifViews/or UI components are touched. - Code quality checks, from
.claude/skills/review/SKILL.md, scoped togit diff develop...HEAD— unconditional, unlike Design compliance.
This is a report-only run: do not append to .claude/context/rejections.md and do not merge — those steps belong to the post-PR /review.
Check 2 — Line-level quality (code-review:code-review)
Run the code-review:code-review skill against git diff develop...HEAD. On an invocation error, don't stall: print ⚠️ code-review:code-review couldn't be invoked here (disable-model-invocation, plugin scope, or unsupported on this agent) — run an equivalent review yourself before /gates. (this check still runs the plugin; /pr-followup uses the built-in code-review, so its warning no longer fits) and continue to the Output format below.
Output format
## Parallel Review — <branch>
### Architecture compliance (Gate 10 + /review checklist)
[✓|✗] <check> — <file:line if failed>
...
Verdict: APPROVED | CHANGES REQUESTED
### Code quality (code-review:code-review)
- <finding> — <file:line> — <severity>
...
(or, on an invocation error: Check 2's warning string)
## Combined verdict
READY FOR /gates | READY FOR /gates (pending your own code-review:code-review pass) | FIX BEFORE /gates: <deduplicated list — same file:line flagged by both checks reported once>Rule: any Check 1 failure (verdict is CHANGES REQUESTED, or Gate 10 fails) forces FIX BEFORE /gates — these are the same rules /gates will enforce as hard blockers. For Check 2, any Critical or High code-review:code-review finding also forces FIX BEFORE /gates; Medium/Low-only findings don't block. Use the middle, qualified verdict only when Check 2 never actually ran (the disable-model-invocation case) and Check 1 is otherwise clean. Only emit the plain READY FOR /gates when both checks ran and are clean (or Medium/Low-only).
Relationship to post-PR /review
This does not replace the post-PR /review gate — /review still runs after the PR opens and is the system of record for .claude/context/rejections.md and the merge decision. /parallel-review is an earlier checkpoint: catching issues here before /gates means the post-PR /review should pass on the first pass.
Done when
Check 1 and Check 2 have both reported (or Check 2's fallback warning was printed instead), one of the three Combined-verdict strings above has been emitted, and any issues found have been fixed.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

