anti-ai-slop
Skill15 starsChanged 5 months agoMITView on GitHub
Apply this skill when auditing, writing, or reviewing Rust code to avoid generic "AI slop" patterns: boilerplate structs with no purpose, copy-paste error handling, meaningless variable names, over-engineered abstractions, or hollow documentation. Triggers: "this looks AI-generated", "too much boilerplate", "meaningless names", "over-abstracted", "audit my code", "humanize this", "why is this so verbose", "remove unnecessary abstractions", "fix docs", "anti-pattern review".
13 other projects use a version of this.
Adopted by
Most recently taken up by
- peterbamuhigire/chwezi-dev-enginetheir own version, first seen Nov 15, 2026
- peterbamuhigire/business-plan-skillstheir own version, first seen Nov 15, 2026
- bismillahinspiringquotes-eng/claude-code-kittheir own version, first seen Nov 8, 2026
- tazhate/anti-ai-sloptheir own version, first seen Oct 20, 2026
- peterbamuhigire/proposal-skillstheir own version, first seen Oct 17, 2026
- misbahsy/anti-ai-sloptheir own version, first seen Sep 14, 2026
Other versions13 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 13
- Hermes-brasil/hermes-brasil★ 159same name
- misbahsy/anti-ai-slop★ 90same name
- peterbamuhigire/chwezi-dev-engine★ 31same name
- claude-code-expert/carve-harness★ 16same name
- bismillahinspiringquotes-eng/claude-code-kit★ 14same name
- peterbamuhigire/proposal-skills★ 11same name
- Vinayak-Shukla-03/anti-ai-slop★ 9same name
- peterbamuhigire/business-plan-skills★ 4same name
- willcagas/skills★ 1same name
- tazhate/anti-ai-slop★ 0same name
3 more not listed
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 repository29 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- llms-full.txtllms.txt
- llms.txtllms.txt
- architecture-diagramSkill
- atomic-commitSkill
- build-rustSkill
- codacySkill
- crates-io-name-checkSkill
- dora-reportSkill
- goap-agentSkill
- harnessSkill
- issue-triageSkill
- lint-rustSkill
- metrics-reporterSkill
- privacy-firstSkill
- release-rustSkill
- reviewing-pull-requestsSkill
- secret-lintSkill
- self-fix-loopSkill
- skill-creatorSkill
- skill-evaluatorSkill
- stacked-prsSkill
- task-decompositionSkill
- test-rustSkill
- tokio-performanceSkill
- triz-analysisSkill
- triz-solverSkill
- verify-actionsSkill
The file
We split this skill into the rules it gives an agent
7 rules. 0 are common, 5 are distinctive, and 2 are very specific to this project.
0
Common
Most of the 11 edited versions we read keep these.
5
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
- Distinctive: line 28
Always name the problem before fixing it.
- Distinctive: line 49
Concrete before abstract: Start with one function. Extract trait only when you have 2+ impls.
- Very specific: line 66
Use thiserror for library errors, anyhow for application errors.
- Distinctive: line 67
Every error message must answer: what was I doing, what input failed, what to do next.
- Distinctive: line 120
Docs are contracts: If it can fail or panic, document it.
- Distinctive: line 122
Zero-cost means earned: Don't add generics for flexibility you won't use.
- Very specific: line 137
Run ./scripts/quality-gates.sh --only clippy while iterating and the full gate before committing.
“Common” is measured: kept by at least 6 in 10 of the 11 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
Anti-AI-Slop Skill — Rust Edition
How to Use
- Audit mode — Share code. Run through checklists below. Name every pattern.
- Creation mode — Read "What to do instead" first, then write code avoiding patterns.
- Spot-fix mode — Point to one element. Diagnose it, rewrite it, explain the fix.
Always name the problem before fixing it.
Part 1 — Code Structure Anti-Patterns
| Pattern | What it looks like | Why it's slop |
| Wrapper struct for no reason | struct MyError(String) with no extra behavior | Just use anyhow::Error or the inner type directly |
| Over-engineered trait hierarchy | trait Processor<T: Handler<U>> for one impl | One trait, one impl = delete the trait |
mod utils | mod utils { pub fn helper(...) } | No concept. Scatter into real modules |
mod common | Grab-bag module used by everything | Symptom of no domain model |
| Phantom generics | struct Foo<T> { _marker: PhantomData<T> } with no bounds | If T does nothing, remove the generic |
| Re-export sprawl | pub use inner::* chains 3 levels deep | Forces users to trace invisible paths |
| Builder for 2 fields | FooBuilder::new().name(x).value(y).build() | Just use Foo { name: x, value: y } |
What to Do Instead
- Model the domain. Name types after what they represent, not their role in the code.
- Flat module tree. One level deep is almost always enough.
- Direct construction. Struct literals over builders unless >5 optional fields or invariants.
- Concrete before abstract. Start with one function. Extract trait only when you have 2+ impls.
Part 2 — Error Handling Anti-Patterns
| Pattern | What it looks like | Why it's slop |
.unwrap() in library code | value.parse().unwrap() | Panics in calling code without context |
| `.expect("should not happen") | everywhere | If it can happen, handle it. If not, prove it. |
Box<dyn Error> in all signatures | fn run() -> Result<(), Box<dyn Error>> | Erases error type, forces dynamic dispatch |
| `map_err(|_| "failed") | Error message that says nothing | Include the input, the operation, what failed |
match err { _ => return Err(...) } | Manual error forwarding | Use ? operator |
| Custom error type per function | ParseConfigError, LoadConfigError, ReadConfigError | One ConfigError with variants |
What to Do Instead
- Use
thiserrorfor library errors,anyhowfor application errors. - Every error message must answer: what was I doing, what input failed, what to do next.
- One error enum per module, not per function.
Part 3 — Documentation Anti-Patterns
| Pattern | What it looks like | Why it's slop |
| Restates the name | /// Creates a new Foo. for fn new_foo() | Zero information gain |
# Examples with trivial code | Example that just calls new() | Show a real use case or delete the section |
| Passive voice everywhere | "This function is used to..." | "Creates", "Returns", "Validates" — be direct |
TODO: implement | Left in committed code | Either implement it or open an issue |
// Safety: this is safe | In unsafe blocks | Explain why the invariants hold, not that they do |
What to Do Instead
- First line: one sentence, imperative, what it does for the caller.
# Errorssection: enumerate whatErrvariants are returned and when.# Panicssection: document every panic condition explicitly.# Safetysection inunsafe: state the invariants required, not just "safe".
Part 4 — Naming Anti-Patterns
| Pattern | What it looks like | Fix |
data, info, result as field names | struct Foo { data: Vec<u8> } | Name the content: payload, bytes, records |
Manager, Handler, Processor types | struct DataProcessor | What does it process? InvoiceParser, LogNormalizer |
Single-letter generics beyond T | fn foo<A, B, C, D> | Name the role: Key, Value, Input, Output |
is_true, has_data booleans | if obj.is_true() | Name the predicate: is_authenticated, has_pending_items |
temp, tmp, val, x in non-trivial scope | 20-line function with let tmp = ... | Name what it holds |
Part 5 — Audit Workflow
- Scan — Check all four pattern lists. Flag every match by name.
- Score severity:
- 🔴 Structural — Wrong abstraction level. Requires refactor.
- 🟡 Surface — Wrong name, wrong error message, missing doc section.
- 🟢 Cosmetic — Minor polish.
- Prioritize — Fix structural first. Don't rename fields on a wrong abstraction.
- Rewrite — Provide the specific replacement, not generic advice.
- Explain the why — Name the Rust principle behind each fix.
Positive Doctrine
- Domain names > technical roles. Types are nouns from the problem space.
- Errors are first-class. Design error types with as much care as success types.
- Docs are contracts. If it can fail or panic, document it. No exceptions.
- Flat is better than deep. One layer of abstraction does more than three thin ones.
- Zero-cost means earned. Don't add generics for flexibility you won't use.
Repository Enforcement
This workspace already fails the build for several patterns above, so "fix" means changing the code, not silencing the lint:
clippy::unwrap_usedandclippy::expect_usedare denied in library code; binaries and tests may opt out with a reasoned attribute.clippy::allow_attributesandallow_attributes_without_reasonare denied, so adding#[allow(...)]is itself a failure unless a reason is supplied.#![forbid(unsafe_code)]applies at the crate roots;// Safety:comments are only relevant to the explicitly reviewedunsafesites that remain.missing_docsand rustdoc link checks run in the quality gate.
Run ./scripts/quality-gates.sh --only clippy while iterating and the full
gate before committing.
Rationalizations
| Rationalization | Reality |
| "This pattern is common in production Rust." | Common doesn't mean correct. Evaluate each abstraction on its own merits. |
| "The AI generated reasonable-looking code." | Reasonable-looking code can hide real issues. Audit systematically, not superficially. |
| "It's just a wrapper struct — harmless." | Every abstraction has a cost. If it adds no behavior, it adds confusion. |
Red Flags
- Finding
mod utilsormod commonin the codebase - More than one error type per module
#[doc = "Creates a new Foo."]onfn new_foo()(zero information gain)
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.

