wicked-testing:execution

wicked-testing:execution

The doer. Takes a scenario or test command, runs it, captures everything, writes the ledger entry. Evidence lives under .wicked-testing/evidence/<run-id>/.

When to use

• You have a scenario ready and need a real run with evidence
• You want to run the existing test suite and record the verdict in the ledger
• You're in a crew test phase and need all scenarios executed

How it dispatches

Input	Dispatch
A scenario file	wicked-testing:scenario-executor
Scenario + "give me a verdict" (acceptance-grade)	Route to /wicked-testing:acceptance — the 3-agent isolated pipeline (writer → executor → reviewer).
Scenario + "give me a verdict" (dev-loop, explicit)	wicked-testing:test-designer ONLY if the caller explicitly asks for the fast-path / self-graded loop.
Pre-written plan, "just execute"	wicked-testing:acceptance-test-executor
"run the suite" (no scenario)	Project's native runner; record result
Contract verification	wicked-testing:contract-testing-engineer
"wire CI" / "bootstrap CI for tests"	Emit CI trigger (pre-push / GH Actions step); record nothing until a real run executes

Default posture: verdict requests go to the 3-agent pipeline. test-designer is the dev-loop fast path with known self-grading risk; it is never the default and never used for audit / CI / crew-phase sign-off evidence. See the warning in agents/test-designer.md.

Dispatch block (executable)

Task(
  subagent_type="wicked-testing:scenario-executor",
  prompt="""Execute the scenario file at the path below and capture evidence.

## Scenario Path
{path to scenarios/<name>.md}

## Evidence Directory
.wicked-testing/evidence/{RUN_ID}/

## Instructions
1. Read the scenario via the Read tool.
2. For each step, run the command via Bash with the scenario's timeout
   (enforce via lib/exec-with-timeout.mjs when available — the shell
   fallback chain is `timeout || gtimeout || bare` with a warning log).
3. Capture stdout, stderr, exit code, wall-clock duration per step.
4. Write step-N.json + evidence.json + artifact files into EVIDENCE_DIR.
5. Determine per-step outcome: exit 0 = PASS, non-zero = FAIL, CLI missing = SKIPPED.

Do NOT self-grade qualitative outcomes. For acceptance-grade verdicts
route to /wicked-testing:acceptance instead."""
)

Swap subagent_type per the table above. For a scenario that also needs contract verification, dispatch scenario-executor and contract-testing-engineer in parallel and merge results.

Tier-2 specialists this skill routes to

For specialized execution paths — chaos experiments, load generators, visual baselines, etc. — dispatch the specialist. Each writes its own artifacts to EVIDENCE_DIR and returns an evidence report the skill includes in the run summary:

Trigger	Specialist
Chaos experiment (Toxiproxy / Chaos Mesh / AWS FIS)	wicked-testing:chaos-test-engineer
Load / perf run (k6 / locust / hey)	wicked-testing:load-performance-engineer
Visual regression run (Playwright + pixelmatch)	wicked-testing:visual-regression-engineer
Full user-journey E2E (multi-context Playwright)	wicked-testing:e2e-orchestrator
Component run (RTL + user-event)	wicked-testing:ui-component-test-engineer
Integration run (real services via testcontainers)	wicked-testing:integration-test-engineer
Fuzz / property run (Hypothesis / fast-check / AFL++)	wicked-testing:fuzz-property-engineer
Security run (SAST scan / DAST scan / secrets check)	wicked-testing:security-test-engineer
AI-feature test (prompt-injection / eval harness)	wicked-testing:ai-feature-test-engineer
IaC validation run (terraform validate / opa / checkov)	wicked-testing:iac-test-engineer
Compliance evidence collection (SOC2 / HIPAA controls)	wicked-testing:compliance-test-engineer
Selective-execution — "which tests for this diff"	wicked-testing:test-impact-analyzer

Chaos / load / security-DAST specialists MUST respect the scenario's trust_level frontmatter field. Production-impacting runs require trust_level: production-authorized AND a change-ticket: reference; otherwise the specialist refuses and records SKIP with reason trust-level-insufficient.

Evidence & ledger

• Every run produces a run_id (UUID v4 from DomainStore)
• Artifacts land in .wicked-testing/evidence/<run-id>/
• manifest.json is written per docs/EVIDENCE.md (produced by lib/manifest.mjs)
• The run + verdict are written to the SQLite ledger
• Bus events emitted (when bus present): wicked.testrun.started, wicked.testrun.finished, wicked.evidence.captured, and finally wicked.verdict.recorded

Output

• The run_id + path to manifest.json
• Verdict (PASS / FAIL / N-A / SKIP)
• One-line summary — never a wall of tool output

References

• docs/INTEGRATION.md
• docs/EVIDENCE.md
• agents/scenario-executor.md, agents/test-designer.md, agents/acceptance-test-executor.md