locator-healer·execution·active

Locator Healer

When a locator fails in CI, proposes a replacement based on accessibility tree diff and test history. Cuts maintenance churn by ~35%.

EVANcursorclaude-codetrigger · ci

Locator Healer

When a locator fails in CI, the healer proposes a replacement based on accessibility tree diff and recent DOM history. Cuts maintenance churn by ~35% on suites of 1k+ tests.

When this skill fires

·A CI run fails with a "locator not found" or "strict mode violation" error

·The diff is significant enough to suggest an actual UI change rather than a flake

·An engineer triggers evan heal against a failing test locally

How healing works

Step 1: Confirm it's not a flake

→The healer never repairs after a single failure. It requires the same locator to fail in the same way across at least two CI runs (or once locally + once in CI). Flakes get retried, not healed.

Step 2: Snapshot diff

The healer pulls the accessibility tree snapshot from the failed run and compares it against the last passing run. Three outcomes:

·Element disappeared — the locator is broken because the feature was removed. Healer asks the engineer to confirm the test should be deleted.

·Element renamed — the role is the same, but the accessible name changed (e.g. "Sign in" → "Log in"). Healer proposes the rename.

·Element restructured — the role hierarchy changed. Healer rerunsLocator Strategist on the new structure.

Step 3: Propose, don't apply

→The healer never silently rewrites locators. It proposes a diff, which the engineer reviews and accepts. Silent healing erodes trust in the suite.

Confidence scoring

Each proposal is annotated with a confidence score:

·High — exact role match, only the accessible name changed

·Medium — role match + minor structural change

·Low — role doesn't match anymore; healer is making a best guess

Low-confidence proposals are flagged red and require explicit /accept to apply.

What it produces

A pull request (or local diff) containing:

·The replacement locator

·A comment explaining what changed in the DOM

·A confidence score

·A link to the failing CI run and the snapshot diff

Why ~35%

→On suites we've measured, locator drift accounts for 35–45% of all test maintenance work. The healer doesn't eliminate it, but it converts most maintenance from "investigate and rewrite" into "review and accept" — which is roughly 1/10th the engineer time.

Related · execution

Scenario Synthesizer

scenario-synthesizer

Builds ramp-up, soak, spike, and chaos scenarios from a single SLO definition. Wires them to ARIA's reporting pipeline by default.

Observability Hookup

observability-hookup

Auto-correlates load test runs with APM metrics (Datadog, New Relic, Grafana) so saturation signals — pool exhaustion, GC pauses — show up in the report.

OWASP Probe Library

owasp-probe-library

Curated probes for the OWASP API + Web Top 10. Read-only and non-destructive — safe to run in shared environments and lower production tiers.

Payload Mutator

payload-mutator

Generates targeted attack payloads for injection, deserialization, SSRF, prototype pollution. Backs off on first signal of damage to stay non-destructive.

Evidence Capture

evidence-capture

For every confirmed finding, records the exact request, response, and timing. Output is replayable in seconds for engineering verification.

Adversarial Probes

nora-adversarial-probes

Red-team test suite covering the OWASP LLM Top 10 — prompt injection (direct + indirect), jailbreaks, system-prompt extraction, guardrail bypass, tool-scope escalation, insecure output handling, model DoS. Generates attack variants automatically and scores severity.