Running Tests

Detect the test framework per changed file, run scoped test commands via Bash, and report pass/fail/error counts. Advise on missing test files.

Freedom level: Adaptive — Follow these steps exactly.

When not to use

Does not review test quality — use reviewing-tests. Does not suggest new tests — use code-testing-agent.

Table of Contents

• Step 0: Create Progress Todos
• Step 1: Identify Changed Files
• Step 2: Find Related Test Files
• Step 3: Detect Test Framework
• Step 4: Run Tests
• Step 5: Report Results and Advise on Missing Tests

Step 0: Create Progress Todos

Use TodoWrite to create todos for Steps 1-5 (all status: "pending"). Mark each status: "completed" as you finish.

Step 1: Identify Changed Files

Resolve the list of changed source files using the following priority order:

1. Explicit path — if the user provided a file path in their message, use it directly
2. Git diff — run git diff --name-only HEAD via Bash to get changed files
3. Conversation context — if a file was recently discussed, use it
4. Ask the user — use AskUserQuestion to request a file path

Filter out non-testable files: binary files, lock files (package-lock.json, yarn.lock, Cargo.lock, poetry.lock), generated files (dist/, build/, .next/, __pycache__/), and config-only files (.env, *.config.js unless the framework has config tests).

Empty-state short-circuit: If git diff --name-only HEAD returns empty and no explicit path was given, tell the user no changed files were detected and stop.

Step 2: Find Related Test Files

For each changed file from Step 1, locate the corresponding test file. Load references/test-runner-guide.md Section 1 for naming conventions and search directories.

Produce a resolved pairs table:

If multiple candidates exist, prefer the one closest in the directory tree. Mark files with no test as Not found — they are addressed in Step 5.

Step 3: Detect Test Framework

For each changed source file, detect the test framework by inspecting config files. Load references/test-runner-guide.md Section 2 for detection signals and run commands.

Monorepo tie-breaking: See Section 5 for the nearest-ancestor rule.

If still ambiguous after checking all signals, use AskUserQuestion:

"I found multiple test frameworks in this repository. Which should I use to test <filename>?"

Step 4: Run Tests

Use Bash to run the scoped test command for each detected framework. Load references/test-runner-guide.md Section 2 for command templates.

Scope the run to only the test files resolved in Step 2 — do not run the full suite.

On failure:

• Capture the exit code
• Surface stderr output directly in the report (Step 5)
• Do not retry or modify the test command

Skip test files marked "Not found" in Step 2 — those are handled in Step 5.

Step 5: Report Results and Advise on Missing Tests

Results Table

Output a results table: Test File | Result | Pass | Fail | Error. Load references/test-runner-guide.md Section 3 for result parsing patterns to extract counts. If counts are not parseable, report exit code only (PASS for 0, FAIL for non-zero).

On failure: Display the relevant stderr excerpt below the table.

Missing Test Advisory

For each source file marked "Not found" in Step 2, output a file-level advisory. Load references/test-runner-guide.md Section 4 for per-language advisory templates.

Advisory is file-level only — do not parse function signatures or suggest specific test cases. For detailed test suggestions, direct the user to the reviewing-tests skill.