메뉴
Evidence before claims, always. No completion claims without fresh verification. The final gate before declaring success.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Use when you need to act as an Elite Software Architect (Maestro) to manage complex repositories. It enforces a "Why over How" philosophy, maintains a persistent project memory (Brain), and orchestrates specialized sub-skills through a Plan-Act-Verify lifecycle.
Elite Tier Backend standards, including Vertical Slice Architecture, Zero Trust Security, and High-Performance API protocols.
Design-first methodology. Explore user intent, requirements and design before implementation. Turn ideas into fully formed specs through collaborative dialogue.
Master specialized skill for building 2025/2026-grade browser extensions. Deep expertise in Manifest v3, Service Worker persistence (Alarms, Offscreen API), Side Panel API, and Cross-Browser compatibility.
The Foundation Skill. LLM Firewall + 2025 Security + Cross-Skill Coordination. Use for ALL code output - prevents hallucinations, enforces security, ensures quality.
Systematic debugging methodology with 4-phase process, root cause tracing, and elite observability standards. No fixes without investigation.
| name | verification-mastery |
| description | Evidence before claims, always. No completion claims without fresh verification. The final gate before declaring success. |
<domain_overview>
Philosophy: Claiming work is complete without verification is dishonesty, not efficiency. Evidence before claims, always. EVIDENCE INTEGRITY MANDATE (CRITICAL): Never claim a task is complete based on assumption or past memory. You MUST generate fresh evidence (logs, screenshots, test output) for every claim. AI-generated success reports are untrustworthy without proof. Any completion signal sent without accompanying verification artifacts must be rejected as "Hallucinated Success".
NO COMPLETION CLAIMS WITHOUT FRESH VERIFICATION EVIDENCE
If you haven't run the verification command in this message, you cannot claim it passes. Violating the letter of this rule is violating the spirit of this rule. </domain_overview> <core_workflow>
BEFORE claiming any status or expressing satisfaction:
1. IDENTIFY: What command proves this claim?
2. RUN: Execute the FULL command (fresh, complete)
3. READ: Full output, check exit code, count failures
4. VERIFY: Does output confirm the claim?
- If NO: State actual status with evidence
- If YES: State claim WITH evidence
5. ONLY THEN: Make the claim
Skip any step = lying, not verifying
</core_workflow> <quality_standards>
| Claim | Requires | NOT Sufficient |
|---|---|---|
| "Tests pass" | Test command output: 0 failures | Previous run, "should pass" |
| "Linter clean" | Linter output: 0 errors | Partial check, extrapolation |
| "Build succeeds" | Build command: exit 0 | Linter passing, logs look good |
| "Bug fixed" | Test original symptom: passes | Code changed, assumed fixed |
| "Regression test works" | Red-green cycle verified | Test passes once |
| "Agent completed" | VCS diff shows changes | Agent reports "success" |
| "Requirements met" | Line-by-line checklist | Tests passing |
| "No errors" | Command output reviewed | "I think it's fine" |
<red_flags>
If you catch yourself:
| Excuse | Reality |
|---|---|
| "Should work now" | RUN the verification |
| "I'm confident" | Confidence ≠ evidence |
| "Just this once" | No exceptions |
| "Linter passed" | Linter ≠ compiler |
| "Agent said success" | Verify independently |
| "I'm tired" | Exhaustion ≠ excuse |
| "Partial check is enough" | Partial proves nothing |
| "Different words so rule doesn't apply" | Spirit over letter |
| "It worked before" | Run it NOW |
| "Too slow to run again" | Slow verification > fast lies |
| </red_flags> |
✅ CORRECT:
[Run: npm test]
[Output: 34/34 passing]
"All 34 tests pass."
❌ WRONG:
"Should pass now"
"Looks correct"
"Tests are green" (without running)
✅ CORRECT:
1. Write test → Run (MUST PASS initial state or FAIL for right reason)
2. Break the code → Run (MUST FAIL)
3. Fix → Run (MUST PASS)
❌ WRONG:
"I've written a regression test"
(without red-green verification)
✅ CORRECT:
[Run: npm run build]
[Output: Compiled successfully]
"Build passes."
❌ WRONG:
"Linter passed, so build should work"
(linter doesn't check compilation)
✅ CORRECT:
1. Re-read plan/requirements
2. Create explicit checklist
3. Verify EACH item with evidence
4. Report gaps or confirm completion
❌ WRONG:
"Tests pass, phase complete"
(tests ≠ requirements)
✅ CORRECT:
1. Agent reports success
2. Check VCS diff (git diff, git status)
3. Verify changes are what was requested
4. Report actual state
❌ WRONG:
Trust agent report without verification
</quality_standards> <integration_and_tooling>
When Ralph Wiggum is active, verification gates are MANDATORY at each iteration:
.maestro/ralph.complete AFTER full verification# Tests
npm test
npm run test -- --coverage
# Build
npm run build
# Lint
npm run lint
npx eslint . --ext .ts,.tsx
# Type check
npx tsc --noEmit
# Tests
pytest
pytest --cov=src
# Lint
ruff check .
flake8 .
# Type check
mypy src/
# Git status (uncommitted changes)
git status
# Git diff (what changed)
git diff
# Process exit code (last command)
echo $? # Unix
$LASTEXITCODE # PowerShell
</integration_and_tooling> <reference_and_audit>
ALWAYS before:
From failure analysis: