Agent-Skills.md

Agent Skills: Test Loop Skill

Plan, generate, and heal an executable E2E test suite from approved acceptance criteria (web and mobile).

UncategorizedID: hoangnguyen0403/agent-skills-standard/test-loop

Repository

hoangnguyen0403/agent-skills-standard
HoangNguyen0403License: MIT
567164

Install this agent skill to your local

pnpm dlx add-skill https://github.com/HoangNguyen0403/agent-skills-standard/tree/HEAD/.codex/skills/test-loop

Skill Files

Browse the full folder contents for test-loop.

Download Skill

Loading file tree…

.codex/skills/test-loop/SKILL.md

Skill Metadata

Name
test-loop
Description
"Plan, generate, and heal an executable E2E test suite from approved acceptance criteria (web and mobile)."

Test Loop Skill

[!IMPORTANT] Plan, generate, and heal an executable E2E test suite from approved acceptance criteria (web and mobile).

Optional args: slug=<feature>, ticket=<id/url>, mode=interactive|autonomous|channel, channel=<id>, auto_continue=true|false, profile=business|hybrid|technical.

Instructions

When the user asks to perform this workflow, execute the following steps:

Test Loop Workflow

Goal: Turn approved ACs into an executable, traced E2E suite, and classify any failure as a repair or a real bug instead of leaving it to manual triage.

Steps

  1. Load scope:
    • slug, operator_profile (carried), AC-*, SRS lanes, build target (web/API/mobile build artifact digest/version), platform/OS matrix, permissions/reset readiness per C6, stack testing skills.
    • Driver preflight for each lane (quality-engineering-playwright-cli / quality-engineering-appium-mcp scripts/preflight.sh); record unavailable drivers in driver_blocked[] (STLC Stage 4).
  2. Plan (STLC Stage 2-3, V-Model System Level):
    • Run specialist-test-planner per quality-engineering-test-plan-authoring to produce test_plan_path and selector_gaps using C1/C2/C6 templates and skills/quality-engineering/quality-engineering-test-plan-authoring/references/lifecycle-and-roles.md.
    • BLOCKED (no stable AC-* trace) if trace missing; BLOCKED (HALT: <trigger>) when planner triggers HALT.
  3. Prepare selectors:
    • Run specialist-testid-inserter on selector_gaps; in interactive mode stop for approval on its APPROVAL: required list; in autonomous mode pass approved_production_edits only when granted, else carry gaps forward in selector_gaps_remaining[].
    • For lane: web, build/extend page objects per quality-engineering-playwright-pom-generation into page_objects[]. Mobile uses stable accessibility IDs per quality-engineering-selector-stability.
  4. Generate:
    • One scenario per specialist-integration-test-generator call, seed-first, via lane driver (web: playwright-cli/Playwright MCP; mobile: Appium MCP/platform runner). Unstable locator scenarios land in selector_gaps_remaining[].
  5. Run and heal (STLC Stage 5: Triage, Flaky Analysis, Defect Drafting):
    • Run generated suite; per failure, run specialist-test-healer into heal_verdicts[].
    • HEALED only with RERUNS: 3/3 green, ASSERTION_DELTA: none, and fixture/selector integrity preserved; unchanged assertion text does not authorize weakening fixtures or widening selectors. REAL_BUG_DO_NOT_HEAL logs to real_bugs[] (defect drafting for dev-fix); QUARANTINE_CANDIDATE routes to quality-engineering-flaky-triage into flake_quarantine[]; BLOCKED (missing driver/evidence) recorded in missing_evidence.
    • Visual baseline changes require reviewed diff with named approver per quality-engineering-visual-baseline.
  6. Handoff (STLC Stage 6 Closure):
    • Calculate Automation Health (quality-engineering-automation-health), carry release_confidence, open defects, and residual risks per skills/quality-engineering/quality-engineering-test-plan-authoring/references/lifecycle-and-roles.md. Technical PASS validates test suite only; route to verify-work.

Runtime Contract

  • Use after implement-feature reaches GREEN, or whenever ACs have E2E/mobile lanes without executable coverage.
  • Required inputs: slug, stable AC-* trace, a runnable build/app target.
  • Return BLOCKED (no build target or AC trace) only when the build target cannot be established or AC-* is missing.

Handoff Payload

  • slug, operator_profile, test_plan_path, assumed_results[], halt_triggers[], page_objects[], generated_tests[], driver_blocked[], heal_verdicts[], flake_quarantine[], selector_gaps_remaining[], real_bugs[], release_confidence, outcome report, next workflow.

Blocking Questions

  • Ask max 3 at a time with a recommended default and 2-3 options.

Output Template

# Test Loop Report: [Name]
## Scope
## Plan
## Generated Tests
## Page Objects
## Selector Gaps Remaining
## Heal Verdicts
## Flake Quarantine
## Real Bugs Found
## Automation Health
feedback_loop_minutes: ; suite_reliability_pct: ; release_cadence: ; prod_escape_rate: ; release_confidence: high | medium | low
## Outcome Report
{schema_version: 1, run_id: "[run-id]", slug: "[slug]", workflow: test-loop, feature_status: implemented, started_at: "[timestamp]", completed_at: "[timestamp]", requirement_trace: {brd_objectives: [], requirements: [], acceptance_criteria: [], srs: []}, completed_evidence: [], missing_evidence: [], decision_needed: [], recommended_next_workflow: verify-work, cost: {source: unavailable}, agent: {identity: "[agent-identity]", model: "[model]"}}
## Next Workflow
verify-work | dev-fix
## Cost Report
Call `get_session_cost(workflow="test-loop")` before final handoff.