ralph-harness

Canonical rules for the external-reviewer bootstrap, liveness heartbeat freshness gate, and convergence detection. This skill is the single source of truth for reviewer-warmup behavior.

Use only when the user explicitly invokes `$ralphharness`, requests RalphHarness in Codex, asks RalphHarness to handle a named phase, or explicitly requests autonomous or quick mode or continuation without pauses.

Load this skill suite for any spec that involves end-to-end testing, browser automation, Playwright, MCP browser tools, VE tasks, user flow verification, UI testing, or integration tests that drive a real browser. Covers session lifecycle, environment setup, navigation anti-patterns, stable-state detection, auth flows, selector stability, and cleanup guarantees.

This skill should be used only when the user explicitly asks to use `$ralphharness-design`, or explicitly asks RalphHarness in Codex to run the design phase.

This skill should be used only when the user explicitly asks to use `$ralphharness-design`, or explicitly asks RalphHarness in Codex to run the design phase.

This skill should be used only when the user explicitly asks to use `$ralphharness-implement`, or explicitly asks RalphHarness in Codex to run implementation for approved tasks, quick mode, or an explicit continue request.

This skill should be used only when the user explicitly asks to use `$ralphharness-refactor`, or explicitly asks RalphHarness in Codex to revise spec artifacts after implementation learnings.

This skill should be used only when the user explicitly asks to use `$ralphharness-requirements`, or explicitly asks RalphHarness in Codex to run the requirements phase.

This skill should be used only when the user explicitly asks to use `$ralphharness-research`, or explicitly asks RalphHarness in Codex to run the research phase.

This skill should be used when the user asks to rollback a spec execution to the pre-execution git checkpoint.

Use only when the user explicitly invokes `$ralphharness`, requests RalphHarness in Codex, asks RalphHarness to handle a named phase, or explicitly requests autonomous or quick mode or continuation without pauses.

This skill should be used only when the user explicitly asks to use `$ralphharness-start`, or explicitly asks RalphHarness in Codex to start or resume a spec.

This skill should be used only when the user explicitly asks to use `$ralphharness-tasks`, or explicitly asks RalphHarness in Codex to run the tasks phase.

This skill should be used only when the user explicitly asks to use `$ralphharness-triage`, or explicitly asks RalphHarness in Codex to triage a large effort into multiple specs.

This skill should be used when the user asks to "build a feature", "create a spec", "start spec-driven development", "run research phase", "generate requirements", "create design", "plan tasks", "implement spec", "check spec status", "triage a feature", "create an epic", "decompose a large feature", or needs guidance on spec-driven development workflow, phase ordering, or epic orchestration.

This skill should be used when the user asks about "ralph arguments", "quick mode", "commit spec", "max iterations", "ralph state file", "execution modes", "ralph loop", "coordinator behavior", "delegate to subagent", or needs guidance on Ralph plugin arguments, state management, delegation patterns, or execution loop behavior. Core behavioral skill for all RalphHarness operations.

This skill should be used only when the user explicitly asks to use `$ralphharness-cancel`, or explicitly asks RalphHarness in Codex to stop execution or remove a spec.

This skill should be used only when the user explicitly asks to use `$ralphharness-feedback`, or explicitly asks RalphHarness in Codex to draft or submit feedback.

This skill should be used only when the user explicitly asks to use `$ralphharness-help`, or explicitly asks RalphHarness in Codex for help or command guidance.

This skill should be used only when the user explicitly asks to use `$ralphharness-implement`, or explicitly asks RalphHarness in Codex to run implementation for approved tasks, quick mode, or an explicit continue request.

This skill should be used only when the user explicitly asks to use `$ralphharness-index`, or explicitly asks RalphHarness in Codex to generate or refresh index artifacts.

This skill should be used only when the user explicitly asks to use `$ralphharness-refactor`, or explicitly asks RalphHarness in Codex to revise spec artifacts after implementation learnings.

This skill should be used only when the user explicitly asks to use `$ralphharness-requirements`, or explicitly asks RalphHarness in Codex to run the requirements phase.

This skill should be used only when the user explicitly asks to use `$ralphharness-research`, or explicitly asks RalphHarness in Codex to run the research phase.

This skill should be used only when the user explicitly asks to use `$ralphharness-start`, or explicitly asks RalphHarness in Codex to start or resume a spec.

This skill should be used only when the user explicitly asks to use `$ralphharness-status`, or explicitly asks RalphHarness in Codex for status or active spec progress.

This skill should be used only when the user explicitly asks to use `$ralphharness-switch`, or explicitly asks RalphHarness in Codex to switch the active spec.

This skill should be used only when the user explicitly asks to use `$ralphharness-tasks`, or explicitly asks RalphHarness in Codex to run the tasks phase.

This skill should be used only when the user explicitly asks to use `$ralphharness-triage`, or explicitly asks RalphHarness in Codex to triage a large effort into multiple specs.

Execute deterministic code quality validations as a quality gate for RalphHarness task execution. Runs Layer 3A smoke test (Tier A AST, <1 min), Layer 1 (test execution), Layer 2 (test quality analysis), Layer 3B (Tier B BMAD Party Mode, ~15 min). Uses Two-Tier approach: Tier A (AST deterministic) + Tier B (BMAD Party Mode consensus). Generates a checkpoint JSON consumed by RalphHarness VERIFY steps. Use when you need to validate that code meets quality standards before COMMIT.

Core Smart Ralph skill defining common arguments, execution modes, and shared behaviors across all Ralph plugins.

This skill must ALWAYS be invoked at the start of every spec — it audits the agent's own system prompt for broken references before any work begins. Invoke unconditionally regardless of goal keywords. Detects phantom infrastructure, ghost paths, incorrect URLs, missing CLI tools, and absent .env files referenced in CLAUDE.md, copilot-instructions.md, or any active system prompt instructions.

This skill should be used when the user asks to "verify a fix", "reproduce failure", "diagnose issue", "check BEFORE/AFTER state", "VF task", "reality check", "check test quality", "mock-only tests", or needs guidance on verifying fixes by reproducing failures before and after implementation, or detecting mock-heavy test anti-patterns.