git:20260220.518720a to git:20261010.496d6f4

2 added, 0 removed. Audit A to A.

---

name: revyl-mcp-analyze

description: Analyze failed Revyl MCP test executions and classify them as real bugs, flaky tests, infra issues, or test-design improvements.

+ metadata:

+ internal: true

---

# Revyl MCP Analyze Skill

## Evidence Flow

1. Start from task ID after `run_test`.

2. Pull status payload:

- `get_test_status(task_id="...")`

3. Inspect failed step expectations and observed reasoning.

4. Classify:

- REAL BUG

- FLAKY TEST

- INFRA ISSUE

- TEST IMPROVEMENT

5. Recommend exact next tool/action.

## Classification Rules

- REAL BUG: actions complete but app behavior contradicts expected outcome.

- FLAKY TEST: app behavior is acceptable but validation is too brittle.

- INFRA ISSUE: no meaningful app steps executed or setup/device/build failure.

- TEST IMPROVEMENT: test structure is weak or allows ambiguity.

## Required Output

```text

Task: <task-id>

Classification: <REAL BUG | FLAKY TEST | INFRA ISSUE | TEST IMPROVEMENT>

Confidence: <HIGH | MEDIUM | LOW>

Evidence:

- Expected: <summary>

- Observed: <summary>

- Why: <short rationale>

Next action:

- <specific update_test / rerun / bug filing action>

```