adr-writerlisted

# Skill: ADR Writer > **Load trigger:** `"load adr-writer skill"` > **DORA:** Cap 3 (AI-Accessible Internal Data) > **Token cost:** Low ## Purpose Produce Architecture Decision Records (ADRs) in a format that: - Humans can understand and approve - Future agents can read as authoritative context - Links decisions to DORA capabilities so the Dojo can reference them ## When to Write an ADR Write an ADR when a decision: - Affects the architecture layer structure or dependency direction - Adds or removes a mandatory tool or CI gate - Changes the agent routing strategy - Establishes a convention that agents will follow going forward - Supersedes an existing ADR Do NOT write an ADR for: - Implementation choices within a single file - Temporary workarounds (use a code comment + TODO issue instead) - Decisions the team hasn't made yet (ADRs record decisions, not options) ## Numbering Convention ADRs live in `docs/adr/`. Number sequentially: `ADR-001`, `ADR-002`, etc. Query existing ADRs: `ls docs/adr/` before assigning a number. ## Full ADR Template ```markdown # ADR-[NNN]: [Short Imperative Title] > **Status:** Proposed | Accepted | Deprecated | Superseded by ADR-[NNN] > **Date:** [YYYY-MM-DD] > **Author:** [Human name or "AI-assisted, approved by [name]"] > **DORA Capability:** Cap [N] — [Name] > **Dojo Module:** [Belt] — [Module name] (if applicable) ## Context [1–2 paragraphs. What situation forced this decision? What constraints exist? What problem are we solving?