test-suite

name	test-suite
description	Delegate an ad-hoc test run to Codex and report results in chat (no artifact) — the cheap inner-loop step; the code-review agent is the checkpoint. Use on /test-suite or 'run the tests'.

/test-suite — Run tests, report in chat

Ad-hoc test runs. Detects the runner per .agents/testing.md, runs the suite once, and reports results in chat. Writes no artifact.

This is the cheap inner-loop step: run it between commits while iterating to confirm tests stay green, then run the code-review agent at a checkpoint for the full six-axis review (which runs tests itself as its gate and embeds the result in the .reviews/ artifact). Use /test-suite when you want to know whether tests pass without paying for a full review.

Delegate to Codex

codex exec -C "$PWD" "/test-suite"

Run directly in Claude only if codex is unavailable or the user asks.

Process

1. Detect the runner per .agents/testing.md (package.json scripts.test → bun test/vitest/jest; pyproject → pytest; Cargo.toml → cargo test; go.mod → go test ./...; etc.). If multiple, prefer the one CI runs.
2. Install deps if needed (only if unresolved — no node_modules/.venv). Prefer bun install per global §5.
3. Run once — capture stdout+stderr, exit code, wall time. No retries.
4. Report in chat:

**Runner:** <bun test / pytest / ...>
**Command:** <exact command>
**Status:** pass | fail | partial | error
**Counts:** N passed, N failed, N skipped (total N)
**Duration:** X.Xs

<one-line summary; on failure, list test name + key error excerpt>

What NOT to do

• Don't write artifact files. That's the code-review agent's job (the combined .reviews/ artifact).
• Don't edit test files to make them pass. Failing tests become findings.
• Don't mark pass when tests were skipped due to install failures — that's error.
• Don't re-run until green. One run per invocation; flaky tests are findings worth a /ticket.

Activity

After the run, emit a feed event with the result:

# pass:
.claude/bin/activity tests-pass "Tests pass · <N> passed" "<runner> in <time>"
# fail:
.claude/bin/activity tests-fail "Tests fail · <F> failed / <N>" "<runner>"