What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
A configuration file sitting in your repository does not prove that your coding agent loaded it for the task that failed. First verify the product, interface, repository, and file scope; then look for evidence that the file entered the agent’s context. Discovery and instruction-following are separate questions.
Why is my AI coding agent ignoring my instructions?
The cause is often a mismatch between the file you edited and the agent session you are checking: the wrong checkout or working directory, an unrecognized filename, a path rule that does not match, or a more local or higher-priority instruction. Discovery rules vary by product and interface, so do not assume that a filename or location supported in one agent works in another.
As an Amazon Associate I earn from qualifying purchases.
Use this 15-minute sequence as a suggested timebox, not a measured completion time. Reproduce the issue in the same IDE, CLI, cloud-agent task, or chat surface where it occurred.
Run the 15-minute audit
Minutes 0–2: Reproduce the failure precisely
- Record the agent or product and the interface or mode you used.
- Confirm the repository or workspace root and current directory; make sure you are in the same checkout as the failed task.
- Choose a small, repeatable task that shows the behavior, and note the target file or path involved.
Minutes 2–5: Confirm the expected file
- Check the exact filename, spelling, capitalization, and location. Confirm the file exists in this checkout and contains the instruction you expect.
- Verify that the agent recognizes that file type in this particular product surface. A file’s presence alone is not evidence that it is a supported instruction source.
Minutes 5–9: Check scope, precedence, and size
- Inspect instruction files at the repository root and along the path to the target file. Look for nested files, overrides, and user- or organization-level instructions that may affect the result.
- For path-specific rules, confirm that the rule’s glob or other scope actually matches the files in the task.
- Check whether a size limit, profile, or configuration directory changes what the agent reads. The details differ by product; Codex and GitHub Copilot have distinct documented models below.
Minutes 9–12: Seek evidence that the file loaded
Use a status or source-inspection method documented for the product and surface. A source reference or active-instruction listing is useful evidence of discovery, though it does not prove adherence.
#1 Best Overall
Minutes 12–15: Run a discriminating test
Give the agent a narrow, harmless task whose answer depends on one unique instruction in the file. Compare the result with both the instruction and any source evidence. Do not use secrets or risky commands as diagnostic instructions. If the file appears to have loaded but the answer conflicts with it, investigate ambiguity, conflicting rules, and task relevance rather than assuming the file was missed.
How to check Codex instruction discovery
OpenAI’s Codex AGENTS.md documentation describes project instruction files being combined from the repository root downward, with nearer files appearing later in the combined prompt. Check the root-to-task chain for nested AGENTS.md files and AGENTS.override.md, as well as user-level instructions. A more local instruction can therefore affect what the agent receives for a task.
The documented default combined project-instruction limit is 32 KiB. Guidance beyond that limit can be omitted once the limit is reached. The documentation also describes project_doc_max_bytes and configurable fallback filenames: a renamed file is not automatically discovered just because it contains instructions. If you change Codex configuration, restart Codex or run it again so the change can load.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problemsTo verify the session, check Codex’s reported workspace root and ask it to list its active instruction sources. The documentation also describes a TUI log approach. These checks help establish which sources were available; they do not show that every instruction was followed.
How to check GitHub Copilot instruction scope
GitHub documents several instruction-file classes in its repository instructions documentation. Check that the file type matches the intended scope and that any path rule applies to the task’s files.
| Instruction file | Intended scope | What to verify |
|---|---|---|
.github/copilot-instructions.md |
Repository-wide instructions | Confirm it is in the repository’s .github directory and that the Copilot feature or surface you used supports it. |
.github/instructions/NAME.instructions.md |
Path-specific instructions | Check its applyTo glob against the files in the task. A rule that does not match the task’s paths will not apply to them. |
AGENTS.md |
Agent instructions | Check the nearest applicable file; GitHub documents that the nearest file takes precedence. |
Repository-wide and matching path-specific instructions may both apply. In Copilot Chat, inspect the response’s reference list for evidence that a repository instruction file was used. Support varies across Copilot features, so a file’s existence does not establish that every Copilot surface will use it.
Rank #4
What if you use another coding agent?
Do not transfer Codex or Copilot filenames, precedence rules, glob behavior, toggles, or verification methods to another product. Confirm the current documentation for the exact agent, version, and interface you used before deciding that a particular file should load. Discovery behavior can differ by product and surface.
If the file loaded but the agent did not follow it
Loading and adherence are different diagnostic outcomes. Once you have evidence the instruction was available, inspect whether it conflicts with a more local or higher-priority rule, whether it applies to the task, and whether it states an actionable requirement. Then test a clearer version of the instruction on the same small task. A successful load check alone cannot establish that the agent will follow every instruction.
Quick Recap
Best Value
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.
