archlisted

# /mneme:arch — Architecture analysis and decision support Read-only analysis of an architecture question. Draws on recalled mneme memory and the codebase, proposes 2-3 solutions with trade-offs, recommends one. Does NOT modify code. Context comes from mneme (`recall`) plus generic repository anchors (`CLAUDE.md`, `docs/`, `README`) — this skill depends on no project-specific layout. ## Arguments `/mneme:arch "<question>"` — architecture question or decision to analyze. Examples: - `/mneme:arch "how should the auth module be structured?"` - `/mneme:arch "should this service be split in two?"` - `/mneme:arch "REST vs gRPC for the internal API?"` ## Permissions (VIOLATION = ABORT) - Read files: YES - Write / Edit files: FORBIDDEN — this skill analyzes, it never changes code - Bash: FORBIDDEN - Memory recall (`mcp__plugin_mneme_memory__recall`): YES - Memory remember (`mcp__plugin_mneme_memory__remember`): permitted ONLY to STAGE a decision the user explicitly asked to persist — never automatically, and it stages for review, it does not publish "Read-only" here means read-only with respect to the codebase: no file is created or edited. Staging a memory via `remember` is the one permitted side effect, and only on user request. ## Procedure ### Step 0: Recall prior memory — MANDATORY Before researching the codebase, call `mcp__plugin_mneme_memory__recall` with the architecture question (key concepts, modules, technologies). 1. There is NO scoring/judge step — mneme