Why can Claude Code not find my test command?

Because nothing has told it what your test command is. Without a CLAUDE.md it infers one from your package manifest or from convention, and a repo using a non-default runner, a monorepo script or a Makefile target will not match that guess.

Claude Code says the tests passed but I do not think it ran them. How do I check?

Ask for the command it ran and the last lines of its output. A real test run has a count in it. If you get a summary with no command and no numbers, treat it as not run. This is the most common way a change ships unverified.

How do I tell Claude Code what my test command is?

Put it in CLAUDE.md under a Commands heading, written exactly as you would type it, flags included. Then add a matching rule to permissions.allow in .claude/settings.json so it can run without asking every time.

What if some tests genuinely cannot run locally?

Say so in CLAUDE.md — name the command, and name what it needs that a laptop does not have. A documented gap gets reported honestly. An undocumented one gets silently skipped and read as a pass.

Claude Code cannot find my test command

The prompt is minor. What it leads to is not: a step that could not run gets read as a step that passed.

Last updated: 2026-09-20

Why the prompt is the small part

A missing test command looks like a minor annoyance. The damage is what comes next: a step that could not run is easy to report as a step that did not fail. You get a green summary describing a test run that never happened.

A failing test is information. A test that never ran, reported as a pass, is worse than no test at all — because now you believe something.

The fix, both halves

1. Tell it the command. In CLAUDE.md at the project root:

## Commands
- Test: `pnpm vitest run --silent`
- Lint: `pnpm eslint .`

Verbatim. Including the flags. It is copied, not interpreted.

2. Let it run without asking. In .claude/settings.json:

{ "permissions": { "allow": ["Bash(pnpm vitest:*)", "Bash(pnpm eslint:*)"] } }

Without the second half you have traded a "cannot find it" prompt for an "may I run it" prompt.

How to tell whether tests actually ran

Ask for	Good answer	Bad answer
The command	The exact string, matching your repo	A paraphrase, or nothing
The output	A count: "42 passed, 0 failed"	"Tests pass"
The exit code	`0`, and it is stated	Not mentioned

Make it a habit to ask. It costs one line and it is the difference between a verified change and a claimed one.

Watch for the flags that fake a pass

Some test runners will report success when they ran nothing at all. --passWithNoTests does exactly what it says. So does appending || true to a command. If either appears in a command that is supposed to be proving your change works, the proof is gone.

Worth writing into your CLAUDE.md never-list, where it applies to every session automatically.

Claude Code forgets my project every session is the same file solving the bigger version of this problem. Claude Code keeps asking permission is the second half of the fix above.

What does it cost to have someone set this up?

Flat prices, written scope, stop at any rung. No call to book, no retainer.

Rung	What you get	Clock
Free — starter repo	A real setup you copy yourself	now
$29 — Setup Lite	One project: CLAUDE.md, allowlist, one skill, as a pull request you review	24 hours
$99 — Setup Sprint	3–5 skills, hooks, up to 3 MCP servers, plus a runbook	48 hours
$79 — Surface Audit	Already set up and still fighting it: contradictions and an exact fix list	48 hours

Setup Lite $29 Or start free

Clocks run from a complete intake, not from the Stripe receipt.

Why the prompt is the small part

The fix, both halves

How to tell whether tests actually ran

Watch for the flags that fake a pass

Related

What does it cost to have someone set this up?