/architecture-viewer

architecture-viewer

Generate the two architecture-review deliverables for a target repo: the JSSI-branded interactive HTML viewer and the Findings/Open-Decisions register.

Usage

/architecture-viewer <target-repo-path> [--silent]

$ARGUMENTS is the path to the target repository to review, optionally followed by --silent (or phrasing like "run silent" / "in silent mode") to run the full pipeline with zero AskUserQuestion calls.

What it does

Invokes the av-orchestrator skill against the supplied repo path.

• Default (interactive) mode: reads the repo (delegating to jssi-legacy-comprehension:system-map, dependency-boundary-mapper, user-flow-mapper, resilience-mapper, performance-mapper, and contract-inventory-mapper), builds the 16-tab diagram set, edits a copy of the approved reference viewer template to produce docs/architecture/viewer.html, and writes docs/architecture/index.md. May ask one focused clarifying question via AskUserQuestion for genuine scoping ambiguities (e.g. which service is primary in a monorepo).
• Silent mode (--silent / "run silent" / "silent mode"): runs the identical pipeline start to finish with zero AskUserQuestion calls, makes every scoping decision unaided, and records each one with its rationale in a required "Excluded and why" / "Assumptions made" narrative subsection of index.md.

Every finding is grounded in real file+symbol/line evidence; every gap (missing retry, missing scan, missing dedup key, etc.) is flagged even when nothing is syntactically wrong. Findings (code-level, fixable) and Open Decisions (architecture choices needing a human ruling) are always kept in two physically separate sections and never merged. This command never resolves an Open Decision and never walks a human through the existing Open Decisions table — for that, run /architecture-viewer-interview after both deliverables exist.

This is an internal Tier-1 tool: local read/write only, no network egress, no git operations, no ambient credentials.