review-slop
Skill4 starsChanged 3 months agoMITView on GitHub
Detect slop in a branch or PR — correct work that should not be in the diff. Verbose comments and docs, duplication across sibling files, ceremony without a threat model, scaffolding shipped as product, checks that must be remembered, and verification sitting in the wrong layer. Use when reviewing a change heavy in docs, config, CI or deploy plumbing, or when asked "what can we cut", "is this too verbose", "review for slop".
1 other project uses a version of this.
Adopted by
Most recently taken up by
- Agenta-AI/agentatheir own version, first seen Nov 1, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- Agenta-AI/agenta★ 4.9ksame 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 repository19 other files it gives its agents.
- AGENTS.mdAGENTS.md
- adapter-contract-testingSkill
- azure-devops-prSkill
- bugszero-root-causeSkill
- legacy-3pSkill
- port-refactoringSkill
- quality-gateSkill
- review-ai-writingSkill
- review-allSkill
- review-architectureSkill
- review-complexitySkill
- review-driftSkill
- review-impactSkill
- review-prSkill
- review-securitySkill
- review-testingSkill
- scope-challengeSkill
- teams-local-messagesSkill
- write-testsSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 38
Read the commit messages too: a war story already in one does not also belong in a comment.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Slop Review Skill
The lens no other reviewer here applies: code and prose that are correct and still should not be in the diff.
Reference
docs/slop-review-guide.md — the seven patterns with the
quotes they came from, how to confirm each, and the counter-pressure test. Read it first,
including the counter-pressure section; a reviewer that skips it deletes the checks that
catch silent failure.
Protocol: docs/review-protocol.md — arguments, ref resolution, exhaustiveness, severity levels, report shape. Read it first.
Its § 4 Be exhaustive is mandatory; the lens steps it refers to are steps 3–4 (Apply the slop lens, Apply the counter-pressure) below.
docs/accidental-complexity-guide.md is a
different axis (smells inside domain code). Where a finding is really about a missing value
object or a defensive null check, it belongs to review-complexity, not here.
Process
- Resolve refs and gather the diff — per
review-protocol.md. Read the commit messages too: a war story already in one does not also belong in a comment. - Look outside the diff before judging it. This lens depends on protocol § 4 more than the others do: three of the seven patterns — duplication across siblings, verification in the wrong layer, more than one source of truth — name a change outside the diff, so a diff-only pass finds none of them. Read the whole of every new file, and find the siblings that have no hunk at all.
Apply the slop lens, per pattern:
- Prose that isn't ours — for each comment block and doc paragraph: is it in the tool's own documentation, in the line below it, or in the commit message? Judge by category remaining, never by lines removed.
- Checks that can't fire, or must be remembered — for each assertion, name the runtime state that trips it. For each check, name what invokes it; if the answer is "a human who remembers", say so.
- Verification in the wrong layer — for anything CI does in shell, ask whether the Makefile, a check script or the deploy tool already does it, and whether a laptop can run it.
- More than one source of truth — for each new secret or pipeline variable, what
does it authorise, and is the value already committed? For each fact added to the
agent entry file (
CLAUDE.md,AGENTS.md), is it also indocs/? - Duplication across siblings — for each new file, find the file it was copied from and diff them. For each repeated block or name-set, grep it and count the files. Then enumerate the rest of that family: a block duplicated into four of five siblings is the same finding as one omitted from the fifth.
- Scaffolding kept as product — for each new script or module, who runs it after this branch lands?
- Over-claiming — for each confident claim in prose or in a commit message, run it if it can be run, and quote the result.
- Apply the counter-pressure before writing any finding that removes a check. State what still catches the failure. If nothing does, it is not a finding.
- Confirm § 4 of the protocol is satisfied, then generate report →
reviews/slop-review-issues.md
Severity Model
| Severity | Indicators |
| Critical | A second source of truth for a value that must not drift; a check that gives false assurance (opt-in, or structurally dead); a claim asserted as fact that is inference |
| Major | Verification in a layer only CI can run; scaffolding shipped as product; a block, file or name-set duplicated across siblings |
| Minor | Comments and docs restating public documentation, the code, or the commit message |
Exclusions
- Smells inside domain code — those are
review-complexity. - Style and formatting — owned by the linters.
- "This is verbose" with no named cost. Every finding names what the next maintainer pays.
- Whether content belongs in this repo at all. A runbook for a system the repo does not build may well be in the wrong place, but that is an ownership call for the people who own the work, not a defect a diff reviewer should rule on.
Output Format
Per review-protocol.md § 6, with two additions — every
finding names the cost of keeping it, and every finding that removes a check names what
still catches the failure:
### 1. Title
**Location:** path/to/file:123
**Pattern:** one of the seven
**Issue:** what is there that should not be
**Cost of keeping it:** what the next maintainer pays — drift, false assurance, a second
path to maintain. Not "it is verbose".
**Still caught by:** (checks only) what fails if this goes. "Nothing" means keep it.
**Suggestion:** the specific deletion or move
**Reference:** a file in this repo that already does it right, if there is oneThe summary paragraph says what the change carries that it does not need, and what that costs.
When to run it
On a diff heavy in docs, config, CI or deploy plumbing — where the other lenses have little to say. On a code-only change it usually reports nothing, which is a correct result, not a reason to invent findings.
Example Usage
review-slop # Current branch vs main
review-slop feat/foo # Specific branch
review-slop feat/foo dev # Different base
review-slop 30 # PR/MR by numberDoes 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.

