test

Installation
SKILL.md

Run tests through the twin. The request that accompanied this invocation names a test — a file, a name pattern, or a full command after -- — or names nothing. No .veris/run.sh and no direct-tier .veris/setup.json → stop; setup runs first. A setup.json with "tier": "direct" replaces run.sh: run the flow directly against the wired sandbox and read the trace where a gate reads the receipt (direct.md). Read .veris/setup.json: test_command is the runner the repository uses.

1. What to run

  • Named. Map the reference onto that runner, never onto another: a file → <runner> <file>; a pattern → the runner's own name filter (jest -t, pytest -k, go test -run); text after -- → verbatim.
  • Nothing named. Candidates are test_command, plus every test file that reaches the vendor and is not already in it: it imports the vendor's client or SDK, names its hostname, or lives under integration/e2e. Find them with grep; do not run yet. A file whose only contact with the vendor is a mock is not a candidate. State the list in one line, then run.

2. Run

Installs
2
First Seen
7 days ago
test — veris-ai/plugins