from rlm-workflow52
Orchestrates an RLM-inspired repository workflow with phase gates, lockable artifacts, and automated bootstrap for run folders — enforces TDD, traceability, and
Provides a complete repository workflow for implementing requirements via a strict phased process (Phase 0..7). It enforces worktree isolation, lockable artifacts with SHA-256 locks, TDD discipline, TODO hard gates, and traceability. The skill includes detailed phase contracts, gates, and commands to bootstrap runs and verify locks.
Use when you need deterministic, auditable implementation runs: feature work, bug fixes requiring root-cause analysis, or large refactors that must follow strict gate checks and manual QA sign-off. Triggers include phrases like 'Implement requirement <run-id>' or 'Run RLM Phase <N>'.
Works best with agents that can run repo-local scripts, manage git worktrees, and spawn subagents for parallel phases (Codex, OpenCode, agents with subagent/task tools).
RLM Workflow is a comprehensive repo-level workflow orchestrator enforcing phase gates, TDD, artifact locking with SHA-256 verification, and git worktree isolation. Scripts are well-structured with dual Python/PowerShell implementations and a bash wrapper. Most scripts fail gracefully in isolation (missing references/ dir or no run data) — they expect to run inside a fully bootstrapped repo, not standalone. verify-locks.py ran cleanly. No security concerns detected.
Impressively thorough workflow system with strong guardrails (Iron Laws for TDD, branch protection, lock chains). Scripts are clean, no shell injection risks, no network calls, no telemetry. Only deduction: scripts can't run standalone outside a bootstrapped repo context, and the references/ dir dependency isn't handled gracefully (hard FileNotFoundError). Architecture is exemplary — lean SKILL.md with progressive disclosure to references/, dual-platform scripts, clear output contracts.
