From yolo
Three-tier verification pipeline (quick/standard/deep) with goal-backward methodology, convention checks, anti-pattern scanning, and requirement mapping.
How this skill is triggered — by the user, by Claude, or both
Slash command
/yolo:verification-protocolThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Authoritative spec for YOLO's verification pipeline. The QA agent is deprecated; Dev natively verifies via MCP testing tools.
Authoritative spec for YOLO's verification pipeline. The QA agent is deprecated; Dev natively verifies via MCP testing tools.
/yolo:vibe execute mode (unless turbo)must_haves.artifacts path existscontains value found via grep{placeholder}, TBD, Phase N stubsQuick, plus:
must_haves.key_links confirmed via grepCONVENTIONS.md if it exists (see §5 / VRFY-06)Standard, plus:
| Signal | Tier |
|---|---|
--effort=turbo | Skipped |
--effort=fast | Quick |
--effort=balanced | Standard |
--effort=thorough | Deep |
| >15 requirements or last phase before ship | Deep (override) |
Precedence: context overrides > effort-based > default.
Start from desired outcomes, derive testable conditions, verify against artifacts. Catches code that exists but doesn't fulfill its purpose.
must_haves.truths -- each must be verifiably true in codebasepath, contains each contains string, provides declared capabilityfrom file references to file, pattern matches viaActive when .yolo-planning/codebase/CONVENTIONS.md exists. Silently skipped otherwise.
| Tier | Behavior |
|---|---|
| Quick | Skipped |
| Standard | Spot-check naming patterns + file placement for new files |
| Deep | Systematic: every new file vs naming, every modified file vs conventions, code patterns vs documented idioms |
Categories: naming patterns, file placement, import ordering, export patterns.
| Anti-Pattern | Detection | Severity | Tier |
|---|---|---|---|
| TODO/FIXME without tracking | grep -rn "TODO|FIXME" not linked to tracker | WARN | Deep |
| Placeholder text | {placeholder}, TBD, Phase N stubs, lorem ipsum | FAIL | Standard+ |
| Empty function bodies | Functions with no implementation | FAIL | Deep |
| Filler phrases | "think carefully", "be thorough", "as an AI" in agent/ref files | FAIL | Standard+ |
| Unwired code | Exported symbols never imported elsewhere | WARN | Deep |
| Dead imports | Imported symbols never used | WARN | Deep |
| Hardcoded secrets | sk-, pk_, AKIA, ghp_, glpat-, password/secret patterns | FAIL | Standard+ |
Severities: FAIL = must fix before ship. WARN = review, may be intentional.
Notes: placeholder detection excludes template files; filler detection applies to agent/ref/command files only; secret detection uses pattern matching (known prefixes + common patterns).
Deep tier only. Traces requirement IDs to implementing artifacts.
Scope: current phase only. Cross-phase requirements noted but not flagged.
Protocol instructions in agent definitions (not JS hooks or event handlers).
{type}({scope}): {description}. Check only task-related files staged. Self-check protocol.phase, plan, status, completed) and standard sections (Accomplishments, Task Commits, Files Created/Modified, Deviations). Report issues.---
phase: { phase-id }
tier: { quick|standard|deep }
result: { PASS|FAIL|PARTIAL }
passed: { N }
failed: { N }
total: { N }
date: { YYYY-MM-DD }
---
# Verification: Phase {N}
## Must-Have Checks
| # | Truth/Condition | Status | Evidence |
## Artifact Checks
| Artifact | Exists | Contains | Status |
## Key Link Checks
| From | To | Via | Status |
## Anti-Pattern Scan (standard+)
| Pattern | Found | Location | Severity |
## Requirement Mapping (deep only)
| Requirement | Plan Ref | Artifact Evidence | Status |
## Convention Compliance (standard+, if CONVENTIONS.md)
| Convention | File | Status | Detail |
## Skill-Augmented Checks (if quality skills)
| Skill | Check | Status | Evidence |
## Summary
Tier: / Result: / Passed: N/total / Failed: [list]
Result classification: PASS = all pass (WARNs OK). PARTIAL = some fail but core verified. FAIL = critical checks fail.
npx claudepluginhub slavpetroff/yolo --plugin yoloVerifies completed work with a 3-tier evidence-based process. Validates tests, linting, types, builds exist and pass, plus deep audit for milestones and PRs. Enforces no completion claims without fresh evidence.
Verifies completed implementations for quality via subagents: completeness, test execution, code review, pragmatic review, production readiness, reality assessment. Compiles read-only reports.
Orchestrates verification of completed implementations via subagents for completeness checks, test execution, code review, pragmatic review, production readiness, and reality assessment. Compiles read-only report.