rcalisted

# Root Cause Analysis Systematic technical investigation methodology. The core principle: **build a case from evidence, don't chase a hypothesis.** ## Project and User Configuration Load configuration in this order. **When multiple sources define the same binding, the first one found wins** — stop checking lower sources for that binding: 1. Read `resources/personal-skill.md` if it exists — personal investigation preferences. 2. Read `resources/domain-skill.md` if it exists — project-specific investigation context. 3. Check the project's CLAUDE.md / CLAUDE.local.md for environment-level capability bindings (especially the **observability** capability and the **session-logging** capability). 4. For any capability still unbound, use the defaults specified in this skill (terminal output for logging, generic tool discovery for observability). ## Extension Points ### domain-skill.md Provide project-specific investigation context: - Log access commands (e.g., `docker compose logs -f web`, Heroku log drains) - Infrastructure topology (services, databases, queues, third-party dependencies) - Common failure modes and their diagnostic shortcuts - Observability tool access (APM dashboards, error tracking URLs, metric queries) - Known flaky areas to deprioritize in investigations See `examples/domain-plugins/rails-heroku-rca.md` for a complete worked example (Rails + Heroku, shared-DB review apps, log filtering patterns, and an RCA output/documentation protocol) — copy and adapt it