run-tests

Activate this skill when:

• You say or think phrases like
  • "Running tests to verify changes"
  • "Now I should run tests to verify my changes"
• The user requests test execution, using phrases like:
  • "test"
  • "run tests"
  • "test the changes"
  • "build and test"

IMPORTANT: Follow the workflow steps EXACTLY as written.

---

Skill Context

SKILL_BASE_DIR: ${CLAUDE_SKILL_DIR}

⛔ VERSION CHECK: If SKILL_BASE_DIR above shows literal ${CLAUDE_SKILL_DIR} instead of a real path, halt: "This skill requires Claude Code 2.1.69 or higher."

Node.js Check: !node "${CLAUDE_SKILL_DIR}/../../lib/check-node-version.js"

⛔ HALT if Node.js Check shows ERROR.

---

Skill Workflow Instructions

Workflow Rules

MANDATORY:

• ⛔ NEVER run test or build commands directly — use run-command.js for ALL execution
• ⛔ NEVER re-run tests to inspect output — Read the output file instead
• ⛔ ALL diagnostic information comes from files, never from re-execution
• Use memory to recall test commands for this project

1. Determine How to Build and Test

Check memory, CLAUDE.md, and project files (package.json, pom.xml, go.mod, Cargo.toml, Makefile, etc.) to determine:

• The test command to run
• Whether a separate build step is required (most projects don't need one)
• The output format (TAP, pytest, go, jest, etc.)

Quiet/silent flags: Use --silent, --quiet, or equivalent when the build tool writes structured results to files (e.g., Maven surefire XML reports, TAP reporters, Jest JSON output). Omit quiet flags when the tool's console output is the only source of test results.

If uncertain: Use AskUserQuestion. Save the answer to memory.

2. Build (if needed)

Most projects don't need a separate build step. Skip this step unless you have determined one is required.

Run build command:

node "<SKILL_BASE_DIR>/scripts/run-command.js" --cmd "<build command>" --out dist/build.log

Parse the JSON output. If exitCode is non-zero: → DELEGATE_TO: references/extract-build-errors.md ⛔ READ FILE AND FOLLOW INSTRUCTIONS, THEN RETURN HERE → Show analysis, ask user: "Enter plan mode to implement fixes?"

• Yes → use EnterPlanMode with analysis context
• No → exit workflow