Test — Pick Suites and Run Them

skills/j-test/ is the canonical, hand-edited directory for this skill; the frontmatter name is j.test, never j.j-test. This skill is a user-invoked convenience runner. This file holds only the conversation and the interpretation of results: discovery, de-duplication, rendering, parsing the answer and execution are scripts in skills/j-test/scripts/, whose headers are the contracts. Do not restate or re-implement their logic here.

Boundaries

• It never writes board status (no task, story or epic status) and never replaces the tester agent. A j.test result is information for the user, not a verdict on any board item.
• The developer agent never invokes it on its own: the developer never runs tests (see CLAUDE.md).
• Per-scope test-run policy (which run must pass before a task, story or epic closes) is a separate mechanism owned by the tester agent: see scope-policy.sh and scope-test-gate.sh in skills/j-test/scripts/ and the "Scope test policy" section of project/documentation/test-runner.md.
• Blank input means all suites: running the full suite is the default.

Instructions

Resolve the scripts directory once, and create one scratch file for the catalog:

TEST_DIR="$([ -d skills/j-test/scripts ] && echo skills/j-test/scripts || echo node_modules/@jenga-ai/agent/skills/j-test/scripts)"
CATALOG="$(mktemp)"

Shell variables do not survive between separate commands: define TEST_DIR again in each later command, and use the literal catalog path that mktemp printed.

1. Build the catalog, once

Tell the user discovery can take up to a minute on a large repository, then run:

bash "$TEST_DIR/build-catalog.sh" > "$CATALOG"

Reuse $CATALOG for every later step. Never run discovery again in this session. If the script exits non-zero, show its stderr and stop.

2. Nothing runnable

If jq '.runnable | length' "$CATALOG" is 0, say so plainly, show bash "$TEST_DIR/build-catalog.sh" --render "$CATALOG" (it lists any not_runnable entries and warnings) and stop. Do not invent a command.

3. Show the list

Print bash "$TEST_DIR/build-catalog.sh" --render "$CATALOG" verbatim, including its final prompt line, and take the user's answer exactly as typed (blank included).

4. Parse the answer

bash "$TEST_DIR/parse-selection.sh" <N> "<answer>"

<N> is jq '.runnable | length' "$CATALOG". Exit 0: its stdout is the suite numbers, in order. Exit 1: show its stderr messages and ask again, repeating until the answer is valid or the user cancels; nothing runs on an invalid answer. Never interpret the answer in this file.

5. Run

bash "$TEST_DIR/run-suites.sh" --catalog "$CATALOG" <numbers...>

Stream its output as it runs and relay its Test Results summary unchanged.

6. Interpret and finish

By run-suites.sh's exit code, which is also this skill's own result:

• 0: every selected suite passed; say so.

• 1: name the failed suites, then offer next steps:

  What would you like to do?
  1. Re-run only the failed suites (step 5 again, with their catalog numbers)
  2. Stop here
  3. Other (describe below)

• 2: validation or usage error, nothing ran; show its stderr.
• 130: interrupted; say which suites did not finish.

When the user is done (including after a re-run), remove $CATALOG.