root-fix-gate
Skill4 starsChanged 31 days agoMITView on GitHub
Use 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.
In this repository47 other files it gives its agents.
- AGENTS.mdAGENTS.md
- anti-slop-engineeringSkill
- anti-slop-engineeringSkill
- api-ergonomics-and-contract-governorSkill
- apple-craft-and-frame-budgetSkill
- apple-craft-and-interaction-physicsSkill
- asset-provenance-guardSkill
- backprop-guardSkill
- chaos-and-degradation-choreographerSkill
- chromatic-drift-guardSkill
- cognitive-load-and-attention-governorSkill
- comment-budget-guardSkill
- context-window-and-token-economics-architectSkill
- cryptographic-agility-and-secret-lifecycle-gateSkill
- decision-guardSkill
- demock-guardSkill
- design-system-drift-guardSkill
- deterministic-simulation-and-property-testingSkill
- distributed-consensus-and-concurrency-gateSkill
- finops-and-unit-economics-governorSkill
- future-scale-gateSkill
- grumpy-senior-staff-engineerSkill
- honeypot-trap-guardSkill
- house-tell-guardSkill
- install-anti-slopSkill
- llm-eval-and-prompt-injection-fortressSkill
- mandela-guardSkill
- mechanical-sympathy-and-perf-gateSkill
- mechanical-sympathy-profilerSkill
- motion-runtime-guardSkill
- observability-and-slo-architectSkill
- observability-and-slo-gateSkill
- paranoid-principal-architectSkill
- pr-boundary-guardSkill
- privacy-and-data-lineage-gateSkill
- re0-handoff-guardSkill
- reduction-guardSkill
- repo-memory-and-invariant-stewardSkill
- safe-evolution-and-rollout-choreographerSkill
- scientific-fault-isolationSkill
- styleseed-guardSkill
- supply-chain-and-sbom-sentinelSkill
- tagore-guardSkill
- type-level-and-ast-slop-gateSkill
- ui-and-copy-disciplineSkill
- verification-oracle-architectSkill
- zero-downtime-migration-gateSkill
The file
We split this skill into the rules it gives an agent
7 rules. About 0 are common, 7 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈7
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 31
Do not turn every CRUD edit into an architecture review.
- Distinctive: line 38
Falsify the elegant hypothesis: Test whether a one-model fix breaks auth, persistence, backward compatibility, ordering, privacy, or …
- Distinctive: line 40
Plan the causal diff: Every changed path needs a reason: owner, consumer, test, migration, or integration.
- Distinctive: line 47
[Trace](usecases/trace.md): Read in full when deciding which layer owns a repeated symptom or data rule, before editing.
- Distinctive: line 48
[Review](usecases/review.md): Read in full after implementation or when reviewing a wide diff, before declaring completion.
- Distinctive: line 71
[Ownership model](references/ownership-model.md): Read in full for MVC, cross-cutting rules, legacy code, and identity-versus-order …
- Distinctive: line 72
[Record contract](references/record-contract.md): Read before saving or auditing an ownership decision packet.
≈ “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
Root-Fix Gate
Fix the right responsibility at the lowest coherent boundary; prove why the change surface exists.
Why This Skill Is Needed & Target State
An agent can make every screen pass while leaving the invalid contract in place: hiding the same user email in twenty controllers instead of checking whether email belongs to identity or checkout. Correct-looking output, passing local tests, a smaller file count, and the user's suggested fix are each insufficient to prove a sound design.
The target is a causal ownership decision before code, followed by focused implementation and an independent diff-to-plan review. The outcome may be a one-boundary correction, a genuinely necessary multi-layer change, pause for missing evidence, or stop when no change is required.
Activation and boundaries
| Signal | Action |
| Repeated filtering, masking, normalization, or validation across consumers | Trace where the rule should be owned before duplicating it |
| Small request produces a cross-layer or unexpectedly wide diff | Compare source-of-truth correction with distributed patches |
| Agent or user says “just change the model” without contract evidence | Examine actual boundaries and downstream dependencies |
| Old fixes keep returning as new symptoms | Reopen the causal hypothesis with failed-probe evidence |
| Owner known and change is narrow, reversible, fully specified | Use parent Anti-Slop BUILD/FIX directly; do not require a record |
| New stack, vendor, or long-term architecture is undecided | Route to decision-guard first, then return for implementation ownership |
Do not turn every CRUD edit into an architecture review. Do not insist that every rule belongs in an ORM model: ownership may be a domain service, schema constraint, serializer, policy layer, transport adapter, or the UI depending on the invariant.
Invariant Rules
- Goal is not a patch instruction. Restate observable behavior separately from the proposed code modification.
- Trace writes as well as reads. A presentation symptom can originate from a data contract, mapping, default, workflow, or authorization rule.
- Owner before consumers. Locate the component responsible for enforcing the invariant. Consumers may still need changes for compatibility.
- Falsify the elegant hypothesis. Test whether a one-model fix breaks auth, persistence, backward compatibility, ordering, privacy, or another caller.
- No forced centralization. Distinct domain contexts can intentionally own different email policies. A global interceptor is not automatically correct.
- Plan the causal diff. Every changed path needs a reason: owner, consumer, test, migration, or integration. File count alone is never a blocker.
- Preserve authority and data. Changing nullability does not authorize deleting existing emails, rewriting identities, or running migrations.
- Tests prove contracts, not formatting. Regression evidence must protect the invariant and affected consumers; avoid tests that merely assert the workaround runs.
- Machine review is not an architecture judge. Automated checks can identify missing evidence and unexpected paths, not certify causal truth.
Execution Workflows
- Trace: Read in full when deciding which layer owns a repeated symptom or data rule, before editing.
- Review: Read in full after implementation or when reviewing a wide diff, before declaring completion.
Decision router
| Evidence | Decision | Next step |
| Root boundary observed, no material dependent risk | proceed | Test that boundary; implement the smallest coherent change |
| Root boundary plausible but auth/data/consumer effects unknown | pause on risky mutation | Read callers, add characterization probes, ask only for inaccessible facts |
| Different contexts have distinct valid owners | proceed with scoped owners | Preserve separate invariants; reject a fake shared utility |
| Root correction changes an authorized contract | pause for authority | State old/new contract, migration impact and user decision needed |
| Original request is already satisfied by a source or standard capability | stop | Show evidence and avoid unnecessary code |
| Diff is wider than expected | review, not automatic rejection | Reconcile paths against evidence and the change plan |
Mandatory checkpoints
Before editing: Name the symptom, the suspected cause, the responsible owner, source evidence, a plausible alternative (including local patches), invariants, and expected changed paths. For a narrow verified fix, a short note suffices. For consequential cross-layer work, use the record contract.
At implementation: Apply the parent BUILD/FIX red-green-refactor cycle. No schema migration, account-email deletion, external send, deployment, or destructive action without its separate authorization.
After implementation: Compare actual paths with the plan; inspect repeated transformations and whether the owner is still wrong. Reopen the hypothesis if the diff spreads without explanation or a regression reveals another owner.
Domain References
- Ownership model: Read in full for MVC, cross-cutting rules, legacy code, and identity-versus-order examples.
- Record contract: Read before saving or auditing an ownership decision packet.
For a persistent packet, run python3 scripts/root_fix_guard.py audit <record.json> --json from the Anti-Slop repository, or its package CLI equivalent npx anti-slop-engineering root-fix audit <record.json> --json only in a release containing the command. Add --base <ref> --repo <project> to compare a real working diff. A standalone copy of this Skill does not include the repository-level script: if unavailable, review manually and report machine audit: NOT RUN. Exit 0 means structurally complete, 2 incomplete, 1 invalid input or Git inspection; exit 0 never proves the root cause is correct.
Failure taxonomy
| Failure | Signal | Recovery |
| Symptom fan-out | Same transformation in many consumers | Trace upstream owner and test a single-boundary fix |
| Model magnetism | Every policy forced into an ORM entity | Check domain, adapter and identity boundaries independently |
| Domain conflation | Order contact email treated as login identity | Separate purposes and test migration/auth contracts |
| False minimalism | One-line change silently breaks dependent callers | Expand the necessary test/migration surface |
| Diff laundering | Unplanned paths explained after the fact | Compare original plan, sources and actual changes |
| Authority bypass | “Clean up” deletes historical user data | Pause, request explicit decision and recovery plan |
| Verification theater | Only UI-hidden symptom is tested | Test source invariant and dependent behaviors |
Delegated Skills
anti-slop-engineering: Own BUILD/FIX, test integrity, SRE and diff verification after ownership is identified.decision-guard: Own independent solution necessity and consequential architecture choices; this Skill owns implementation responsibility.future-scale-gate: Evaluate whether an architectural boundary redesign introduces irreversible evolution constraints.
Completion and limits
Report the chosen owner with source, alternative rejected and why, planned versus actual paths, protected invariants and test evidence, outstanding uncertainty, and proceed/pause/stop. Do not promise that fewer files or a structural pass is a root-cause proof. Semantic scenarios in evals/scenarios.json are a test bank, not executed model-performance evidence.
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.

