Recommended Free Tools
To make coding agents respect a repository’s architecture, combine three things: instructions written for the harness they actually use, automated checks for rules that can be tested, and a small trial in that harness to confirm the instructions are discovered and followed. Prose explains intent and trade-offs; lint rules and structural tests catch repeatable violations.
Start with instructions the agent can discover
Before writing architectural guidance, identify the coding harness and its current instruction mechanism. A rule in a file the harness does not load cannot guide the agent. Visual Studio Code’s documentation lists AGENTS.md as an instruction option for OpenAI Codex and describes other project-wide and targeted instruction formats across supported tools. Check the current documentation for the harness you use rather than assuming filenames or discovery behavior are interchangeable: Visual Studio Code: Configure AI for your codebase.
As an Amazon Associate I earn from qualifying purchases.
Make the project-level guidance concrete enough to orient a change. Include the architecture and boundaries that matter, important directories, established conventions, build and test commands, and what a completed change must satisfy. Visual Studio Code’s guide recommends covering these areas. Prefer a short, prioritized set of repository decisions over a catalogue of preferences: explain the allowed dependency directions and the checks that demonstrate a change fits.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Clear out junk files and repair common Windows errors3Fix the driver behind crashes, sound loss and screen glitchesScope rules to the code that needs them
Keep truly universal constraints in repository-wide guidance. If a subsystem has different conventions or boundaries, attach the rule to that area using the mechanism supported by the harness. Mixing unrelated local rules into a global instruction file can obscure the constraints that matter for any given edit.
#1 Best Overall
| Harness or use | Documented instruction mechanism | Scope |
|---|---|---|
| GitHub Copilot in Visual Studio Code | .github/instructions/**/*.instructions.md files with applyTo patterns, as described by Visual Studio Code’s codebase configuration guide |
Targeted to files matching the patterns |
| OpenAI Codex | Root and nested AGENTS.md files, as described by Visual Studio Code’s codebase configuration guide and custom instructions documentation |
Directory-based; instructions are discovered from the repository root down to the working directory |
| Copilot code review | .github/copilot-instructions.md for repository-wide review guidance, root AGENTS.md for project context, and .github/instructions/**/*.instructions.md for path-specific review guidance, according to GitHub’s code review documentation |
Repository-wide or path-specific, depending on the file |
These mechanisms are not interchangeable guarantees. In particular, when testing nested Codex guidance, use the relevant subdirectory as the working folder; the discovered instruction context depends on the root-to-working-directory path described in the Visual Studio Code guide.
Turn critical boundaries into executable checks
Instructions are useful for explaining why a boundary exists and how a change should fit it. For a critical rule that can be expressed mechanically, add a custom linter or structural test so violations are caught repeatably instead of relying solely on an agent to interpret prose. OpenAI describes this approach in “Harness engineering: leveraging Codex in an agent-first world”: “In practice, we enforce these rules with custom linters and structural tests, plus a small set of ‘taste invariants.’”
Rank #2
- State the rule and rationale. Describe the boundary in the project instructions, including what belongs where and why.
- Encode the checkable portion. Add a lint rule or structural test for violations that can be identified from repository structure or code.
- Explain the repair. Make failure output tell the agent what is wrong and what an allowed fix looks like. OpenAI notes that custom lint messages can inject remediation instructions into agent context.
- Keep judgment where it belongs. Leave context-dependent architectural trade-offs to review; a check should enforce a repeatable constraint, not pretend every design decision is mechanical.
A check that only says “failed” is less actionable than one that names the violated boundary and points to the permitted direction of change. The goal is not to automate all architecture, but to make the important, deterministic rules hard to miss.
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 reinstallCrashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteVerify discovery and behavior in the intended harness
After adding or changing instructions, verify both that the harness loads them and that the resulting guidance is useful for the targeted work. Visual Studio Code recommends reviewing the pattern, testing instructions in a new chat with the same harness, and asking for a small change to a file matched by the rule. Its guide puts it plainly: “Review the pattern and test the instructions by asking the agent to make a small change to a specific matching file.”
Rank #3
- Check the instruction filename, location, and any path pattern against the chosen harness’s current documentation.
- Start a fresh chat or session with that same harness so the test reflects ordinary instruction loading.
- Request a small, bounded change in a file the rule is meant to cover. Check whether the agent observes the relevant architectural constraint.
- For nested Codex instructions, open the corresponding subdirectory as the working folder when testing.
- Run the repository’s actual lint and structural checks on the resulting change.
If the targeted rule has no effect, investigate whether it was discovered and whether its scope matches the file before adding more prose. A failed automated check should lead to a specific repair; a missed instruction may instead indicate a loading or scope problem.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Keep the enforcement layers distinct
Use project guidance for architecture, rationale, conventions, and context that helps an agent choose among valid approaches. Use lint rules and structural tests for repeatable constraints with clear pass-or-fail conditions. Use the intended harness to confirm that instructions reach the agent, then use human review for choices that depend on context. This separation makes the rules understandable without treating prose as a guarantee or tests as a substitute for architectural judgment.
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.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →
