imMamdouhaboammar/anti-slop-engineering / skills/root-fix-gate
root-fix-gate
VerifiedUse when a coding agent is about to patch the same symptom across controllers, views, endpoints or files; a small behavior change causes a wide diff; a workaround may hide a wrong model, schema or domain boundary; or a user proposes a shortcut with uncertain downstream effects. Trace the causal owner and compare a coherent root correction against distributed patches before implementation or review. Skip narrow, localized fixes with an already established owner.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 8.2 KB |
| agents/openai.yaml | config | 326 B |
| evals/scenarios.json | config | 6.7 KB |
| examples/email-checkout-record.json | config | 1.6 KB |
| references/ownership-model.md | markdown | 4.8 KB |
| references/record-contract.md | markdown | 2.7 KB |
| usecases/review.md | markdown | 2.0 KB |
| usecases/trace.md | markdown | 3.7 KB |
references/ownership-model.md
View on GitHub# Ownership and causal boundaries
## The responsibility test
The right fix is located where the relevant invariant becomes authoritative, not necessarily where a wrong value is visible. An invariant may concern identity, order contact, billing, transport representation, permission, or presentation. A shared string name such as `email` does not prove two contexts have the same responsibility.
| Symptom | First likely owner to inspect | Counterexample that defeats the shortcut |
| --- | --- | --- |
| Same validation repeated in controllers | Domain validation/schema or shared application policy | Each endpoint applies a genuinely different rule |
| Email shown on multiple profile screens | Identity data contract and response serialization | Email is valid identity data but visibility is a presentation policy |
| Email required only when ordering | Checkout order contact boundary and identity constraints | Login, reset, or provider identity requires email |
| Same retry copied through clients | Network/client abstraction or central retry policy | Operations have different idempotency/rate-limit constraints |
| Permission filtering duplicated across routes | Authorization policy close to protected operations | Tenant-specific resources require distinct policy rules |
| UI displays inconsistent totals | Canonical aggregation/data definition | Currency/timezone semantics differ by consumer |
| Legacy adapter emits wrong shape | Adapter/serialization boundary | Core domain contract itself is invalid |
Trace **origin → writes → persistence → transformations → reads → consumers**. Find the earliest location whose invariant is actually wrong. Replacing only the output may conceal the cause; changing an earlier component without checking later contracts may also break the system.
## Choosing the owner
Use five questions in order:
1. What behavior would be false if all UI/controllers disappeared? That rule likely belongs deeper than presentation.
2. Can any authorized entry point bypass the proposed enforcement? If yes, a controller-only check may be insufficient.
3. Would all callers agree on the same rule? If not, a global helper or model hook may be incorrect.
4. Is the data rule a persistence invariant, a domain rule, a workflow policy, or a representation concern? Keep each at its proper boundary.
5. What system contract would disprove the chosen fix? Run that probe before implementing destructive or broad changes.
## What the smallest coherent change means
Count logical responsibilities, not files. Updating a canonical schema, its migration, API contract, three consumer adaptations, and tests can be a better design than one dangerous middleware interception. Conversely, repeated response hacks across many endpoints often point to a missed common rule.
A new abstraction is justified only when it owns a meaningful shared contract. Do not extract merely because two functions mention email. Distinguish intentional duplication in tests from duplicated production invariants. Avoid relocating a rule into an ORM entity when doing so creates side effects on reads, unexpected network calls, or cross-context coupling.
## Evidence levels and escalation
**Observed** means inspected repository source, executed query, official contract, or reproduced behavior. **Derived** means a conclusion reproducibly follows from observed inputs. **Inferred** means a hypothesis, not a verified owner. **Unavailable** means access or environment prevented the check; it never means the path is absent.
For account identity, payment, privacy, authorization, schema migration and retention, missing evidence may require a pause even if implementation looks easy. Delete/migrate production data only with the user's specific authority, a tested recovery path, and the contract impact understood. For a routine reversible local edit with a known owner, do not manufacture a record or elaborate architecture debate.
## Architectural counterexamples
- **Global model masking:** A model getter strips email everywhere. It also strips it from order receipts and identity recovery. The correct fix may be an output serializer for a single public surface, not the model.
- **Per-view suppression:** Twenty screens each hide email because registration improperly requires it. If verified, a corrected identity/checkout contract can remove the need for duplicated patches, while migrations and auth tests remain necessary.
- **Distributed policy by design:** Different tenants have distinct verified retention requirements. A single global normalizer breaks those rules; scoped domain policies are correct even when multiple files change.
- **False root-cause certainty:** A failing API response might come from a stale cache, wrong serialization, or storage invariant. Treat each as a hypothesis until a read/write trace identifies the first incorrect boundary.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Unverified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |