architecture-scanlisted

<!-- Source: https://github.com/xxthunder/xxthunder-agentic-skills/tree/develop/plugins/xxthunder-dev-skills/skills/architecture-scan --> # Architecture Scan Reads the codebase, derives the structure that is actually there, and compares it against what the architecture document claims. Two outputs from one capability: - **No document yet** → a *bootstrap proposal* covering the fixed slots. - **A populated document** → a *drift report* naming where the two disagree. Bootstrap and drift detection are the same operation with a different baseline — bootstrap is a drift report against an empty document — so this is one skill, not two. ## This skill is read-only **It never writes.** Not the architecture document, not an ADR, not the ADR index. It produces a proposal or a report; applying anything routes through `design-record`, which is the record's only writer. That means recording a structural change is a two-step move here — scan, then apply — while `design-record` records a *decision* in one. The asymmetry is deliberate. Recording a decision happens constantly and should be one move; a bootstrap or a drift sweep is occasional and produces a large body of inferred content that deserves reading before it lands. An explicit apply step is a review gate, not friction. Do not "optimise" this by giving the scan write access. One skill owning the format is what keeps the conventions from diverging. ## When This Skill Triggers | Trigger phrase | Operation | |----------------|-