Instalar com Codex ou Claude Copie este prompt, cole no Codex, Claude ou outro assistente e deixe que ele revise a página da skill e instale para você.
Um comando direto ignora o prompt de revisão. Verifique a origem antes de executá-lo.
npx skills add https://github.com/idealic-ai/engine --skill test
O comando permanece em uma só linha. Role horizontalmente para revisá-lo antes de copiar.
Prefere uma cópia local? Baixe os arquivos disponíveis atualmente no SkillsMP.
On selection: Read the corresponding modes/{mode}.md file (e.g., modes/coverage.md, modes/hardening.md, modes/integration.md).
On "Custom": Read ALL 3 named mode files first (modes/coverage.md, modes/hardening.md, modes/integration.md) for context, then read modes/custom.md. The user types their framing. Parse it into role/goal/mindset. Use Coverage's topic lists as defaults.
Record: Store the selected mode. It configures:
Phase 0 role (from mode file)
Phase 1 interrogation topics (from mode file)
Phase 3 walk-through config (from mode file)
1. Strategy (Planning + Interrogation)
Before writing code, use the Anti-Fragile Checklist to generate high-value scenarios.
§CMD_REPORT_INTENT:
1: Interrogating ___ testing assumptions, then building the plan. ___.
Focus: ___.
Not: ___.
§CMD_EXECUTE_PHASE_STEPS(1.*)
The Question Bank (20 Questions for Coverage)
Data Integrity
"What happens if we pass null, undefined, or NaN to strict methods?"
"Can we corrupt the state by calling public methods in the wrong order?"
"Does this data structure maintain its invariants after 1000 mutations?"
"What is the 'Zero State' (empty lists) behavior?"
"What is the 'Max State' (arrays with 10k items) behavior?"
Async & Concurrency
6. "What if the Promise rejects immediately?"
7. "What if the Promise hangs forever?"
8. "What if stop() is called while start() is pending?"
9. "Are there race conditions between UI events and background events?"
10. "Is this function re-entrant?"
Refactoring & Isolation
11. "Can we extract the logic to test it without mocking complex dependencies?"
12. "Are we testing implementation details (private state) or behavior (public API)?"
13. "Is this test brittle? Will it break if we rename a variable?"
14. "Can we use a Factory to simplify test setup?"
15. "Are we over-mocking? Can we use real data objects?"
Domain Specific
16. "What happens during a gap or missing data in the pipeline?"
17. "Does backwards traversal / undo handle state reset correctly?"
18. "Do we handle format or version mismatches?"
19. "What if the input payload is shorter/smaller than expected?"
20. "Does the system recover gracefully from transient failures?"
Interrogation Topics (Testing)
The mode file topics (from modes/{mode}.md) are your primary source. These standard topics are available for all modes as supplementary material. Adapt to the task -- skip irrelevant ones, invent new ones as needed.
Standard topics (typically covered once):
Testing strategy -- unit vs integration vs e2e, test runner, framework conventions
Coverage goals -- what percentage, which modules, critical paths vs nice-to-have
Edge cases & boundaries -- null/empty states, max values, type mismatches, error paths
Mocking approach -- what to mock vs use real, mock libraries, fixture patterns
Data integrity -- state corruption, invariant violations, concurrent mutations
Integration boundaries -- external service contracts, API shape validation, DB queries
Test maintenance -- naming conventions, shared fixtures, test organization, cleanup
Repeatable topics (can be selected any number of times):
Followup -- Clarify or revisit answers from previous rounds
Devil's advocate -- Challenge assumptions and decisions made so far
What-if scenarios -- Explore hypotheticals, edge cases, and alternative futures
Deep dive -- Drill into a specific topic from a previous round in much more detail
Plan Creation
After interrogation completes:
Draft: Use the Question Bank + interrogation answers to brainstorm multiple perspectives (Data, Async, Domain), then execute §CMD_WRITE_FROM_TEMPLATE (Schema: TESTING_PLAN.md).
Expand: Immediately propose 5 additional testing avenues inspired by these different perspectives.
Refine: Ask the user which to include, then update TESTING_PLAN.md.
Walk-through (optional):
§CMD_WALK_THROUGH_RESULTS Configuration:
mode: "plan"
gateQuestion: "Testing plan is ready. Walk through the scenarios before executing?"
debriefFile: "TESTING_PLAN.md"
planQuestions:
- "Any concerns about this test scenario's approach or complexity?"
- "Should the scope change -- expand, narrow, or split this scenario?"
- "Dependencies or risks I'm missing?"
2. Execution
Gateway: select inline, agent handoff, or parallel handoff.
§CMD_REPORT_INTENT:
2: Selecting execution path for test implementation. ___.
Focus: ___.
Not: ___.
Run: Verify it fails as expected (red) or passes (green).
Log: Update TESTING_LOG.md with your status.
Tick: Mark [x] in TESTING_PLAN.md.
2.B. Agent Handoff
Hand off to a single autonomous agent.
§CMD_EXECUTE_PHASE_STEPS(2.B.*)
§CMD_HANDOFF_TO_AGENT with:
{"agentName":"builder","startAtPhase":"2.A: Testing Loop","planOrDirective":"[sessionDir]/TESTING_PLAN.md","logFile":"TESTING_LOG.md","taskSummary":"Execute the testing plan: [brief description from taskSummary]"}
2.C. Parallel Agent Handoff
Hand off to multiple autonomous agents in parallel.
§CMD_EXECUTE_PHASE_STEPS(2.C.*)
§CMD_PARALLEL_HANDOFF with the testing plan split across agents.
§CMD_WALK_THROUGH_RESULTS Configuration:
(uses Walk-Through Config from the loaded mode file: modes/{mode}.md)
Testing Tip:
To verify log output, use logger.getHistory(). This returns the structured log entries (including placeholders) which allows for robust assertion of error conditions without relying on fragile string matching.
