- Backend: FastAPI app with all models, services, schemas, routes - Admin Panel: Next.js dashboard with all modules (CRM, POS, repairs, warranties, billing, notifications, etc.) - Ecommerce: storefront (submodule) + product catalog - AI Agents: .agents/ skill library for Antigravity IDE - Docs: architecture, billing, notification gateway markdown - Submodules: Backend/OpenWA, Ecommerce/File 2/storefront
8.5 KiB
| auto_execution_mode | description |
|---|---|
| 0 | Comprehensive review for bugs, security, code quality, standards, and business-flow integration |
Comprehensive Review
You are a senior ReloPortal engineer performing a read-only review of the requested files or current diff. Review the change through all four lenses below, then produce one prioritized findings report:
- Correctness and security — logic errors, edge cases, nullability, races, resource leaks, API contract violations, security issues, caching mistakes, and improper error handling.
- Code quality — React 18, AntD 5, Redux Toolkit, TypeScript strictness, DSA/complexity, clean code, side-effect hygiene, and reusable components/utilities.
- ReloPortal standards —
.devin/rules/rules.md,.devin/agents.md, and relevant.devin/context/guidance, including forms, services, styling, state, data mapping, imports, and testing conventions. - Business-flow integration — real user journeys, cross-page/cross-slice consumers, routing and access gates, payload/response propagation, loading/error/empty states, and sub-service/tab behavior.
Do not make code changes. Report only concrete findings supported by the diff, repository context, Graphify relationships, or established project standards. Route pure code-style findings to the code-quality category and pure flow findings to the flow category rather than duplicating them.
1. Scope the review
- If the user supplied file paths, review those files and their relevant consumers.
- Otherwise review the current diff using
git diff --name-only HEADandgit diff --stat HEAD. - Review complete target files, not only diff hunks.
- Skip generated files, snapshots, lockfiles, and pure JSON fixtures unless explicitly requested.
- Group changed files by feature and identify the user journeys each group touches.
2. Establish context before reading downstream code
Run Graphify before reading related files:
graphify query "what is <FileName> and where is it used"
graphify affected "<FileName>"
graphify query "what does <FileName> depend on"
Use uvx --from graphifyy graphify when the global command is unavailable. Also query Intentgraph for relevant feature ownership/history when the change is feature-specific.
Build a bounded map of:
- Entry points, callers, downstream consumers, and shared components.
- Services, endpoints, hooks, Redux slices, selectors, and utilities involved.
- Routes, roles, supplier types, tabs, and sub-services reachable from the change.
- Form config → submit handler → service call → response hydration paths.
If Graphify is insufficient, use targeted repository search for the relevant symbol only.
3. Correctness and security checklist
- Logic errors, incorrect conditions, and unreachable branches.
- Null/undefined handling and invalid input edge cases.
- Race conditions, stale async responses, missing cancellation, and resource leaks.
- API request/response shape, endpoint, date, pagination, and status contract violations.
- Incorrect cache keys, stale data, invalidation, or ineffective caching.
- Authentication, authorization, access-control, sensitive-data exposure, and unsafe user input handling.
- Errors are surfaced correctly and are not swallowed.
- Loading, empty, error, and retry states remain correct.
4. Code-quality checklist
DSA and complexity
- Nested
find,includes,indexOf, orfilterlookups causing avoidable O(n·m) work. - Repeated filter/map chains on hot or large collections.
- Sorting inside render or loops.
- Inefficient lookup patterns that should use
MaporSet. - Recursive logic without base-case or memoization safeguards.
- Input, prop, or Redux-state mutation outside slice reducers.
- Independent async calls that should use
Promise.all.
React, AntD, and Redux
- Incorrect
useEffectdependencies or missing cleanup. - Derived state stored unnecessarily, unstable keys, components defined during render, or ineffective/overused memoization.
- Controlled/uncontrolled form conflicts, incorrect AntD component usage, missing
rowKey, pagination, or large-table protections. - Redux logic in components, non-memoized selectors returning new objects/arrays, redundant server-data copies, or unnecessary global state.
TypeScript and clean code
- Explicit or implicit
any, wide/unsafe casts, non-null assertions without proven invariants, or repeated structural types. - Functions over roughly 40 lines, nesting deeper than three levels, magic values, dead code, misleading names, or duplicated logic.
console.log, silent catches, or comments that merely narrate implementation.- Custom date, string, number, currency, or status helpers that should use
src/utils/. - Custom renderers that should reuse existing components such as
TableTag.
5. ReloPortal standards checklist
- No raw Axios; use
requestsorrequestsMsfromsrc/services/index.ts. - Endpoints come from
src/resources/constants.tsx. - Service functions are typed and do not contain unnecessary try/catch handling.
- Forms use
ShadowFormWrapper,className="stdForm",uiVersion="v2", co-located config, and the standard action footer. - Shared state uses Redux only when cross-page; local state belongs in hooks/components.
- API data is not normalized or reshaped unnecessarily; no fallback chains or invented identity values.
- Utility classes and
colors.scssvariables are used instead of inline styles or hex colors. - Imports follow project ordering; no commented-out code or
console.logremains. - Existing shared components and utilities are reused before creating custom equivalents.
6. Business-flow integration checklist
- Previously reachable routes, tabs, sections, and sub-services still work.
- Forms prefill the same fields and preserve add/edit/save/cancel behavior.
- Role-based, supplier-type, and layout visibility remains correct.
- No optional/required field changed silently.
- Request payloads still match backend expectations and response fields still match every consumer.
- Shared component prop changes are reflected at all call sites.
- Redux slice changes are reflected in thunks, selectors, and subscribers.
- Home Sale sub-service, Listing Contract, VDI, Supplier, Payroll, and TransfereeProfile integrations are checked when applicable.
- New access gates or complex visibility logic have an explicit business rationale.
7. Categorize findings
Use these severities:
- Blocker — breaks a user journey, loses data, exposes sensitive data, or makes the change unusable.
- Critical — correctness/security bug likely to fail in production.
- High — significant regression, API mismatch, complexity, or maintainability risk.
- Medium — clear but non-urgent improvement or fragile integration.
- Low — minor convention, readability, or style issue.
8. Output
## Comprehensive Review Summary
**Files reviewed**: <count>
**Diff / branch**: <branch or sha range>
**Journeys touched**: <feature journeys>
**Overall assessment**: <safe / safe after fixes / needs rework>
### Blast radius
- <changed file or symbol> → <consumers and downstream impact>
### Findings (sorted by severity)
| Severity | Lens | Location (file:line) | Finding | Suggested fix |
|----------|------|---------------------|---------|---------------|
| Blocker | Flow | ... | ... | ... |
| Critical | Security | ... | ... | ... |
| High | Quality | ... | ... | ... |
| Medium | Standards | ... | ... | ... |
| Low | Flow | ... | ... | ... |
### Patterns worth standardising
- <concrete cross-file pattern>
### Out of scope / routed findings
- <only concerns outside the requested review scope>
### Verdict
- [ ] Ready as-is
- [ ] Safe after Blocker/Critical/High fixes
- [ ] Needs rework
Every finding must include an absolute file path and line number, a concrete impact, and a specific suggested fix. Do not report speculative issues.
Rules
- Do not fix anything; output findings only.
- Do not make separate reports for each lens; combine all findings into the single output above.
- Use Graphify before reading downstream files.
- Do not duplicate findings across lenses.
- Run
npm run lintornpx tsc --noEmitonly when requested or when validation is necessary to support a finding; report pre-existing failures separately.