Borrowing it
Nothing to install: this file belongs to sbroenne/mcp-windows. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/sbroenne/mcp-windows/main/.github/skills/code-review/SKILL.mdgit clone --depth 1 https://github.com/sbroenne/mcp-windowsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/sbroenne/mcp-windows/code-review)<a href="https://agentmods.dev/skills/sbroenne/mcp-windows/code-review"><img src="https://agentmods.dev/badge/skills/sbroenne/mcp-windows/code-review/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/sbroenne/mcp-windows/code-review"><img src="https://agentmods.dev/badge/skills/sbroenne/mcp-windows/code-review.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00059 | $0.01237 |
| Opus 5.5 | $0.00024 | $0.00495 |
| Sonnet 5.5 | $0.00012 | $0.00247 |
| Haiku 4.5 | $0.00006 | $0.00124 |
Grade A, and why
code-review scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 77 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Review Windows MCP changes
Review for observable correctness and user impact, not preferred coding style. Do not edit product code, merge pull requests, or change repository settings during a review.
Establish the intended behavior
- Read the current PR head, diff, linked issues, and acceptance criteria. Distinguish deliberate breaking changes from accidental regressions; do not demand compatibility that the change explicitly excludes.
- Follow repository instructions and testing guidance. Use the existing implementation and tests to confirm what the public interface actually promises.
- Trace affected callers and result consumers, not just the edited method. Relevant surfaces include tool registration and schemas, raw request validation, CLI parsing and dispatch, shared automation services, batches, tool permissions, result models, recovery hints, help, and examples.
- For re-reviews, inspect existing threads and the complete review summaries, including suppressed findings. Check each against the latest code. Do not repeat a fixed finding or assume that an unresolved thread still represents a bug.
Prioritize these failure modes
| Area | What to verify |
|---|---|
| Input contracts | Omitted, null, empty, whitespace, unknown, and mode-inappropriate arguments remain distinguishable where the contract requires it. A schema alone does not prove the runtime rejects an argument. Check validation before binding erases property presence. |
| MCP/CLI parity | Both entry points reach the same operation behavior. CLI aliases, quoting, child arguments, relative paths, working directories, output, and exit codes preserve intent. Do not infer an alias is invalid without checking the parser. |
| Target identity | An ID resolves the original observed control or fails. Replacement, handle/process reuse, provider failure, owner restart, eviction, and automation-thread disposal must not redirect it through a name or position search. |
| Read and action scope | A failed explicit target must not widen into whole-window text, OCR, another table, or an unrelated dialog. Parent/proximity references and coordinate fallbacks must belong to the intended window and control. |
| Batches and permissions | Validate the entire request before side effects. Dependency checks and execution must parse steps identically. Null steps and invalid property presence are handled. Disabled tools cannot be requested through batch steps or attached snapshots. Only an unambiguous discovery result may supply a later action's reference. |
| Bounded work | Enforce limits before bulk fetching or allocating results, not afterward. Incomplete searches cannot prove absence or uniqueness. Check deadlines, final probes, cancellation, and cleanup without arbitrary timeout padding. |
| Service lifetime | Check caller isolation, pipe permissions, startup coordination, build compatibility, request/output limits, and resource ownership. Whole batches must not interleave; read-only waits must not block their triggering actions. Do not replay requests after an uncertain response. |
| Windows input and outcomes | Verify foreground ownership, focus, current coordinates/DPI, and input return values before acting. Dispatch, dialog disappearance, and an actual application outcome are different evidence. Cleanup must not close user applications or release another operation's resources. |
| Snapshots | Differences must refer to the caller's known baseline and reconstruct the current tree, including actionable IDs. Missing or outdated baselines require a full response. Do not preserve an old ID by assigning it to a replacement control. |
| Measurements and guidance | Compare actual serialized outputs using equivalent captures and metadata. Do not relabel historical benchmarks. Verify examples and recovery hints against current argument names, validation, and MCP/CLI ownership. |
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 1mo ago Changed · +1 tokens per session c24b9e784835
- 1mo ago First seen · 77 lines · 58 tokens per session scan A 74b74867b86e
code-review is a skill published in the GitHub repository sbroenne/mcp-windows (125 stars, last pushed yesterday), licensed MIT. It adds 59 tokens to every session and 1,237 once invoked, about $0.0002 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-14.
Other skills, from other repositories
code-analysis
Use the free built-in .NET SDK analyzers and analysis levels with gradual Roslyn warning promotion. USE FOR: the repo wants first-party .NET analyzers; CI should fail on analyzer warnings; the team needs AnalysisLevel or AnalysisMode guidance. DO NOT USE FOR: third-party analyzer selection by itself; formatting-only…
graphify-dotnet
Use graphify-dotnet to generate codebase knowledge graphs, architecture snapshots, and exportable repository maps from .NET or polyglot source trees, with optional AI-enriched semantic relationships. USE FOR: graphify commands; graph JSON, HTML, SVG, Cypher, Markdown, and Obsidian exports; repository map and…
sonarjs
Use SonarJS-derived rules in .NET repositories that ship JavaScript or TypeScript frontends and need deeper bug-risk, code-smell, or cognitive-complexity checks than a minimal ESLint baseline. USE FOR: SonarQube, SonarCloud, or eslint-plugin-sonarjs setups; frontend code smells; cognitive complexity and deeper…
address-review
Read PR review comments from GitHub, evaluate, address, and reply.
resharper-clt
Use the free official JetBrains ReSharper Command Line Tools for .NET repositories. USE FOR: jb inspectcode; jb cleanupcode; stronger C# inspections, cleanup profiles, and CI-friendly JetBrains analysis. DO NOT USE FOR: replacing tests with inspection output; ad-hoc formatting-only work when the repo intentionally…
complexity
Use free built-in .NET maintainability analyzers and code metrics configuration to find overly complex methods and coupled code. USE FOR: the team wants to find overly complex methods; cyclomatic complexity thresholds are needed in CI; maintainability metrics or coupling thresholds need to be configured. DO NOT USE…