JUnit Failure Fingerprint

Cluster JUnit failures into stable fingerprints so CI triage focuses on root causes, not noisy one-off logs.

Use this skill to compress noisy JUnit failures/errors into repeatable fingerprints.

What this skill does

• Scans one or more JUnit XML files
• Extracts only failing/error test cases
• Normalizes volatile tokens (IDs, numbers, line numbers, addresses, UUIDs)
• Generates stable fingerprint hashes for similar root-cause failures
• Emits grouped triage output (text or json)

Inputs

Optional:

• JUNIT_GLOB (default: test-results/**/*.xml)
• TOP_N (default: 20)
• OUTPUT_FORMAT (text or json, default: text)
• STACK_LINES (default: 3) — number of normalized stack lines to include in fingerprint seed
• FAIL_ON_FAILURES (0 or 1, default: 0) — exit non-zero when any failures/errors are found

Run

Text report:

bash

JUNIT_GLOB='artifacts/junit/**/*.xml' \
TOP_N=15 \
bash skills/junit-failure-fingerprint/scripts/fingerprint-junit.sh

JSON output for CI annotation/upload:

bash

JUNIT_GLOB='artifacts/junit/**/*.xml' \
OUTPUT_FORMAT=json \
FAIL_ON_FAILURES=1 \
bash skills/junit-failure-fingerprint/scripts/fingerprint-junit.sh

Run with bundled fixtures:

bash

JUNIT_GLOB='skills/junit-failure-fingerprint/fixtures/*.xml' \
bash skills/junit-failure-fingerprint/scripts/fingerprint-junit.sh

Output contract

• Exit 0 by default (reporting mode)
• Exit 1 if FAIL_ON_FAILURES=1 and at least one failure/error is found
• In text mode, prints summary + top fingerprints
• In json mode, prints machine-readable groups and per-case detail