diff --git a/.github/copilot-instructions.md b/.github/copilot-instructions.md
new file mode 100644
index 000000000..662931e2a
--- /dev/null
+++ b/.github/copilot-instructions.md
@@ -0,0 +1,59 @@
+---
+description: 'Code review guidelines for GitHub copilot in this project'
+applyTo: '**'
+excludeAgent: ["coding-agent"]
+---
+
+# Code Review Instructions
+
+A change note is required for any pull request which modifies:
+- The structure or layout of the release artifacts.
+- The evaluation performance (memory, execution time) of an existing query.
+- The results of an existing query in any circumstance.
+
+If the pull request only adds new rule queries, a change note is not required.
+Confirm that either a change note is not required or the change note is required and has been added.
+
+For PRs that add new queries or modify existing queries, also consider the following review checklist:
+- Confirm that the output format of shared queries is valid.
+- Have all the relevant rule package description files been checked in?
+- Have you verified that the metadata properties of each new query is set appropriately?
+- Do all the unit tests contain both "COMPLIANT" and "NON_COMPLIANT" cases?
+- Are all the alerts in the expected file annotated as NON_COMPLIANT in the test source file?
+- Are the alert messages properly formatted and consistent with the style guide?
+- Does the query have an appropriate level of in-query comments/documentation?
+- Does the query not reinvent features in the standard library?
+- Can the query be simplified further (not golfed!).
+
+In your review output, list only those checklist items that are not satisfied or are uncertain, but also report any other problems you find outside this checklist; do not mention checklist items that clearly pass.
+
+## Validating tests and .expected files
+
+The test infrastructure for CodeQL that we use in this project involves the creation of a test directory with the following structure:
+- Test root is `some/path/test/path/to/feature` (mirrors `some/path/src/path/to/query`)
+- At least one test `c` or `c++` file, typically named `test.c`/`test.cpp`, with lines annotated `// COMPLIANT` or `// NON_COMPLIANT`
+- A `.ql` file with test query logic, or a `.qlref` file referring to the production query logic
+- A matching `FOO.expected` file to go with each `FOO.ql` or `FOO.qlref`, containing the test query results for the test `c` or `c++` files
+- Note that some test directories simply have a `testref` file, to document that a certain query is tested in a different directory.
+
+As a code reviewer, it is critical to ensure that the results in the `.expected` file match the comments in the test file.
+
+The `.expected` file uses a columnar format:
+- For example, a basic row may look like `| test.cpp:8:22:8:37 | element | message |`.
+- For a query with `select x, "test"`, the columns are | x.getLocation() | x.toString() | "test" |`
+- An alert with placeholders will use `$@` in the message, and have additional `element`/`string` columns for placeholder, e.g. `| test.cpp:8:22:8:37 | ... + ... | Invalid add of $@. | test.cpp:7:5:7:12 | my_var | deprecated variable my_var |`.
+- Remember, there is one `.expected` file for each `.ql` or `.qlref` file.
+- Each `.expected` file will contain the results for all test c/cpp files.
+- The `toString()` format of QL objects is deliberately terse for performance reasons.
+- For certain queries such as "path problems", the results may be grouped into categories via text lines with the category name, e.g. `nodes` and `edges` and `problems`.
+
+Reviewing tests in this style can be tedious and error prone, but fundamental to the effectiveness of our TDD requirements in this project.
+
+When reviewing tests, it is critical to:
+- Check that each `NON_COMPLIANT` case in the test file has a row in the correct `.expected` file referring to the correct location.
+- Check that each row in each `.expected` file has a `NON_COMPLIANT` case in the test file at the correct location.
+- Check that there are no `.expected` rows that refer to test code cases marked as `COMPLIANT`, or with no comment
+- Note that it is OK if the locations of the comment are not precisely aligned with the alert 
+- Check that the alert message and placeholders are accurate and understandable.
+- Consider the "test coverage" of the query, are each of its logical statements effectively exercised  individually, collectively? The test should neither be overly bloated nor under specified.
+- Consider the edge cases of the language itself, will the analysis work in non-trivial cases, are all relevant language concepts tested here? This doesn't need to be exhaustive, but it should be thoughfully thorough.