find-mismatch
Systematic code review focusing on bugs that break at runtime (contract mismatches, logic/async errors, serialization, styling) and JS/TS static analysis via fallow. Use when reviewing code changes for runtime errors, post-implementation checks, finding contract mismatches, or running static analysis via fallow on changed files.
How do I install this agent skill?
npx skills add https://github.com/utarn/engineer-skills --skill find-mismatchIs this agent skill safe to install?
- Gen Agent Trust Hubpass
This skill performs code review and static analysis using the 'fallow' tool. It requires installing an external Node.js package and executing shell commands. As it processes untrusted source code files and has the capability to modify project content, it presents a surface for indirect prompt injection.
- Socketpass
No alerts
- Snykpass
Risk: LOW · No issues
What does this agent skill do?
Find Mismatch
Run static analysis using fallow and fix issues that it suggests. Walk through every file. Focus only on runtime failures/bugs and fix them. Do NOT report style, formatting, performance suggestions, docs/tests, or "consider using X".
1. Process
- JS/TS (Pre-check): Run static analysis.
- Check if installed:
fallow --version. - Install globally if missing:
npm install -g fallow. - Execute audit:
fallow audit --format json --quiet. - Filter findings by git status:
git diff --staged --name-only. - If staged: auto-apply safe fixes, re-run audit to verify.
- If unstaged: report-only (no modification).
- Check if installed:
- Review Checklist: Apply checklist to all changes.
- Report: Format findings using the specified template.
2. Checklist Categories
1. Cross-Boundary Contracts
- Calling mismatch: Check RPC, HTTP, IPC, event name/subject match receiver.
- Param names/types:
userIdvsuser_id, string vs number, flat vs wrapped. - Return type: Field accesses (
result.field) must match return signature.
2. Serialization & Deserialization
- Casing:
snake_casein DB/backend vscamelCasein JSON/frontend. - Optionals: Field expected by receiver but optional in producer.
- Enums/discriminators: Discriminant fields (
type,kind) and variants match. - Base64: Double-encoding layers.
3. Logic Bugs
- Double counting: Counters incremented both in loop/branch and outside.
- Off-by-one: Index, pagination bounds (
page <= totalvspage < total). - Conditions: Redundant/dead flags, unreachable statements, shadowed variables.
4. Property & Method Access
- Null derefs: Method calls or array index access (
items[0].name) without null/length checks. - Optional chaining: Stop check too early (e.g.
user?.address.streetstill crashes ifaddressis null). - This context: Callback bindings in classes.
5. Async & Concurrency
- Race conditions: Shared mutable states, parallel filesystem accesses without coordination.
- Missing
await: Promise ignored, silent errors. - Leaks: File/DB/network connections not closed in error paths.
6. CSS & Styling
- HSL variables: Raw HSL components (
0 0% 100%) instead of functional format (hsl(0 0% 100%)) fail in Tailwind v4 (unlike v3). - Inline themes: Circular references, selector mismatch (class vs data-theme).
7. Placeholder & Stub Code
- Unused declarations/files/dependencies, TODO comments, empty catch blocks.
3. Language-Specific Pitfalls
- Python: Hallucinated library methods (pandas, numpy, requests), sync/asyncio confusion, mutable default arguments (
def f(x=[])), silent indentation drops. - JS/TS: Backend (Node) vs Frontend (Browser) API mix-ups,
anyabuse, type-only import mismatches, barrel file false positives (exported but unused in production, used in tests/libraries). - C++: Raw vs smart pointer mix-ups, missing standard library header
#include, uninitialized variables, buffer overflows. - Rust: Borrow checker lifetime overcomplication,
unwrap()onNone/Err, locking across.awaityield points. - Java / .NET: Hallucinated annotations/attributes, LINQ N+1 queries, Entity Framework model snapshot drift / missing migrations (check model changes under
Entities/againstMigrations/). - Go: WaitGroup/Channel deadlocks, unexported struct fields in JSON, defer in loop, map concurrent access.
- PHP: Loose type comparisons (
==vs===), array vs object access, key undefined.
4. Fallow Finding Mapping & Fixes
| Fallow Type | Checklist Category | Auto-Fix Action |
|---|---|---|
unused_exports | Placeholder & Stub | Remove export (if used locally) or entire declaration. |
unused_files | Placeholder & Stub | Delete file (check dynamic imports first). |
unused_dependencies | Placeholder & Stub | Remove dependency from package.json. |
circular_dependencies | Cross-Boundary | Defer to manual review (causes init bugs). |
complexity/duplication | Logic Bugs | Defer to manual review. |
5. Output Format
Include ## Review Summary header (only if JS/TS project):
## Review Summary
- **Manual review bugs found**: N
- **Fallow auto-fixed**: N findings (staged files only)
- **Fallow non-auto-fixable**: N findings
- **Fallow verdict**: pass / fail
Format individual findings (add [fallow] tag for fallow-sourced bugs):
## Bug #1
- **File**: <path>:<line>
- **Category**: <category>
- **What's wrong**: <description>
- **Runtime effect**: <effect>
- **Fix**: <code>
How can the creator link this skill?
Add the canonical catalog link to the repository README so users can inspect current installs and available audits. The publishing guide covers the complete discovery path.
<a href="https://skillzs.dev/skills/utarn/engineer-skills/find-mismatch">View find-mismatch on skillZs</a>