Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallOutdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchFabien Sanglard’s agent.md is a set of repository instructions for an AI coding assistant—not a Java quality standard or an automatic fix. Its useful idea is to stop repeating stable review feedback and make it available to the coding harness. For a Java or Spring Boot project, adapt those preferences to the project’s formatter, Java baseline, tests, API commitments, and architecture, then review every generated change yourself.
What Sanglard’s agent.md is meant to do
In his August 21, 2026 article, software developer Fabien Sanglard describes returning to LLM coding in January 2026 and using agentic IDEs later that spring. He found himself repeatedly asking an agent to avoid magic numbers, explain itself with a short comment, or use short function names. He moved recurring preferences into a root-level agent.md that his coding harness loaded into its prompt. He reports better output and more time for architecture and design, but still reads and verifies the generated code. Read Sanglard’s article and rules.
As an Amazon Associate I earn from qualifying purchases.
That distinction matters: a repository instruction file can make expectations easier to reuse, but it does not enforce them like a compiler, formatter, test, or static-analysis rule. Sanglard describes personal qualitative experience, not a controlled Java or Spring Boot evaluation. He writes, “While this ‘trick’ has considerably improved the code generated, this is not a magic bullet that lets me avoid reading the code.”
Which rules translate well—and what Java teams should change
Sanglard’s rules are preferences, not universal prescriptions. Translate the intent into the conventions your team already uses rather than pasting them unchanged. For example, a strict rule to keep every method name under 30 characters may clash with a project’s naming standards; a demand to add comments can produce noise if it is not limited to explaining non-obvious decisions.
| Original principle | Java and Spring Boot adaptation |
|---|---|
| Extract meaningful or recurring values; avoid magic numbers. | Use named constants or enums when they clarify meaning or are reused. Keep obvious one-off values inline. Prefer domain names that explain the value rather than generic names such as VALUE. |
Reduce indentation with early returns and continue. |
Use guard clauses where they make control flow clearer. Preserve validation, exception behavior, and transaction boundaries; do not flatten code at the cost of readability. |
| Keep function names short; use enums rather than boolean parameters. | Use names that are concise but descriptive under the project’s Java naming conventions. Replace a boolean parameter with an enum or a clearer API when the choices have meaning; do not add an enum for a simple, unambiguous flag by rote. |
| Use blank lines and brief comments. | Follow the project formatter and comment where the reason or constraint is not evident from the code. Avoid comments that merely restate an expression. |
| Keep members private; use abstraction layers that communicate with immediate neighbors. | Preserve encapsulation and the application’s established architecture. Respect public API and framework extension points; do not create extra layers simply to satisfy a generic abstraction rule. |
| Always use braces; avoid unrelated edits. | Apply the repository’s formatting and style checks consistently, and keep a task’s diff focused. Let existing project conventions and automated checks resolve style questions. |
| For a bug fix, write a test first, see it fail, implement the fix, and see it pass. | Ask for a regression test that demonstrates the bug, then the smallest appropriate fix and a passing test run. Check that the test fails for the intended reason and that the implementation does not weaken existing behavior. |
Before adopting any rule, reconcile it with the Java version, formatter, build, test framework, public compatibility requirements, and Spring application design. Spring Framework’s repository guidance is one example of project-specific JDK and build-check conventions; it is not a template for every Spring Boot codebase. The available Spring Framework AGENTS.md search-result page supports only that narrow point.
How to make repository instructions useful
- Put stable guidance where the harness reads it. Sanglard uses a root-level
agent.mdloaded by his coding harness and mentions compatible file-name symlinks as part of his setup. Harnesses differ, so confirm the expected filename and loading behavior in your own tool rather than assuming it will discover the file. - Write rules as observable actions. “Keep the change scoped to the requested feature” is easier to apply than “write clean code.” Specify the test-first loop, expected checks, and when to introduce a constant or comment.
- State boundaries and exceptions. Tell the agent to preserve public API behavior, follow existing package and layer boundaries, avoid unrelated refactors, and ask before changing architecture or build configuration. Note where a project convention takes precedence over a general preference.
- Use the project’s real verification commands. Include the build and test commands the team actually runs, if appropriate for the harness and task. Do not claim a test passed unless the agent ran it and reported the result; inspect the diff and verify important changes independently.
- Keep the file maintained. Remove obsolete guidance and resolve contradictions when team conventions change. Instructions are most useful when they reflect current practice rather than an aspirational style guide.
A practical instruction might say: “For bug fixes, add or update a regression test first. Confirm it fails for the reported behavior, implement the narrowest fix, then run the relevant tests. Do not claim verification unless you ran the command. Keep edits limited to the requested issue unless a necessary dependency is explained.” That translates Sanglard’s process preference into verifiable steps without pretending the instruction itself guarantees correctness.
Rank #2
When to use a file versus Spring AI Agent Skills
A harness-loaded repository file and Spring AI Agent Skills solve related but different packaging problems. Sanglard’s file is guidance for coding sessions in a compatible harness. Spring’s January 13, 2026 article describes skills for applications built with Spring AI: folders containing Markdown instructions and YAML frontmatter, with optional scripts and reference material. The documented flow registers skill names and descriptions, then loads full instructions when a task matches. Its Java review-skill example checks security, Spring Framework conventions, null risks, readability, and maintainability. See Spring’s Agent Skills article.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errors| Approach | Where it applies | How guidance is loaded | Operational considerations |
|---|---|---|---|
| Repository instruction file | A coding harness that supports and loads the file for work in the repository. | Instructions are supplied to the coding session according to that harness’s behavior. | Check filename and loading behavior; keep repository-specific guidance accurate. The cited material does not establish quality results across harnesses. |
| Spring AI Agent Skills | An application built with Spring AI that uses the documented skills mechanism. | Skill metadata is registered; matching tasks can load the full instructions on demand. | Spring says locally executed scripts run without sandboxing, there is no built-in human approval mechanism, and skills have no built-in versioning. Review scripts and design suitable isolation, approval, and version management. |
Spring’s article lists spring-ai-agent-utils version 0.4.2 and a Spring AI requirement of 2.0.0-M2+ for its example. These are setup details in that January 2026 article, not a statement of the latest releases; check current release documentation before using them.
Manage context and keep human review in the loop
Sanglard’s explanation for diminishing attention to instructions as a conversation grows is context dilution. He reports two responses: start a new session for each feature and ask the harness to reload agent.md when output quality drops. Treat these as his workflow suggestions, not a proven behavior of every model or IDE.
- Start a focused session when moving to a distinct feature or task.
- When the agent stops following repository guidance, check whether the harness loaded the file and ask it to reload the instructions if supported.
- Review the diff for correctness, scope, API compatibility, security, and consistency with the application’s design.
- Run relevant tests and checks yourself or confirm that the reported commands and results are accurate.
An instruction file can make review expectations more visible and reduce repeated prompting. It cannot establish that generated Java code is correct, secure, or appropriate for a Spring application; those judgments still depend on tests and human review.
Quick Recap
Best Value
Rank #4
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Recommended Free Tools
