skill-routing-arbitratorlisted

# Skill Routing Arbitrator When multiple skills could fire on the same prompt, this skill provides deterministic preferences so Claude routes to the right one. ## Priority hierarchy (general) 1. **Most specific language/domain** wins over generic — e.g., `ecc:python-review` over `ecc:code-review` 2. **Active executor** wins over **strategy/workflow doc** — e.g., `ecc:code-review` (runs review) over `engineering:code-review` (strategy) 3. **Current version** wins over **dated snapshot** — e.g., `anthropic-skills:mcp-builder` over `anthropic-skills:mcp-builder-20260419` 4. **Real skill** wins over **legacy shim** — e.g., `ecc:tdd-workflow` over `ecc:tdd` shim 5. **Connected-data-source skill** wins when data is available — e.g., `common-room:account-research` if Common Room MCP is connected, else `sales:account-research` ## Specific cluster rules ### Code review | Situation | Skill | |---|---| | Language known + active review needed | `ecc:<lang>-review` (cpp, python, go, rust, kotlin, flutter) | | PR / multi-agent comprehensive | `ecc:review-pr` | | Local diff review | `ecc:code-review` | | PR triage (lightweight built-in) | `review` | | Security focus on PR | `security-review` | | Codebase security review | `ecc:security-review` | | Strategy / process discussion only | `engineering:code-review` | ### Testing | Situation | Skill | |---|---| | TDD in language X | `ecc:<lang>-test` | | E2E specifically | `ecc:e2e-testing` | | Generate tests from existing file | `ruflo-te