Skip to main content

frontend-architecture

Frontend component architecture review — dispatch the component-architecture-review agent over the frontend component files to catch reusable components that should be extracted, duplicated UI patterns, prop drilling, component-granularity problems, and inconsistent component APIs as a frontend evolves. Use when the user says "review the frontend architecture", "are my components reusable", "is this UI duplicated", "should this be a shared component", "check for prop drilling", or before extracting a component library. Advisory — it recommends, it does not edit.

Source facts

Repository
bdfinst/agentic-dev-team
Last source activity
August 3, 2026 at 16:58
Detected SKILL.md language
English
Stars
295
Forks
36

Install options

The review-first prompt is selected by default. You can switch to a direct command or download a local copy.

Review the source files

Read SKILL.md and any companion files shown by SkillsMP before deciding whether to install.

Showing SKILL.md

SKILL.md
Source instructions · Read-only preview
name
frontend-architecture
description
Frontend component architecture review — dispatch the component-architecture-review agent over the frontend component files to catch reusable components that should be extracted, duplicated UI patterns, prop drilling, component-granularity problems, and inconsistent component APIs as a frontend evolves. Use when the user says "review the frontend architecture", "are my components reusable", "is this UI duplicated", "should this be a shared component", "check for prop drilling", or before extracting a component library. Advisory — it recommends, it does not edit.
argument-hint
[--path <dir>] [--since <ref>] [--all] [--json]
user-invocable
true
allowed-tools
Read, Grep, Glob, Bash(git diff *), Bash(git log *), Skill, Agent
# Frontend Architecture Role: orchestrator. This command scopes the changeset to frontend component files and dispatches the `component-architecture-review` agent, then aggregates one report. It does not review files itself — it coordinates. `/code-review` dispatches the same agent automatically, but only when a frontend component file is newly *added* (`Scope: added-only`, #1733) — merely modifying an existing one does not trigger it there. This command scopes the agent to the frontend files you name (`--path`/`--since`/`--all`), regardless of whether any are newly added — use it for a diff that only modifies existing components, a standalone frontend-only pass (e.g. before extracting a component library), or to scope beyond a diff. `/repo-review` (#1735) covers the third, unconditional case: the whole component inventory, every pass, with no scoping flags at all. This command is executed under orchestrator direction. Dispatch the agent with its `model:`/`effort:` frontmatter as declared — the harness resolves both fields natively before dispatch, per Model/Effort Resolution in `agents/orchestrator.md` (ADR 0026). ## Orchestrator constraints 1. **Advisory only.** Aggregate findings and recommendations. Do not edit component code. Hand actionable fixes to `/apply-fixes` or `/build`. 2. **Scope to the frontend.** Only pass frontend component files to the agent; skip the run with a one-line note if none are in scope. 3. **No double-reporting.** This review owns component *reuse and composition*. Defer accessibility to `a11y-review`, framework reactivity to the framework-reactivity reviewers (`react-reactivity-review`/`vue-reactivity-review`/`angular-reactivity-review`), non-UI logic duplication and SRP/coupling to `structure-review`/`refactor-opportunity-review`, and naming to `naming-review`. When the same line belongs to one of those, drop it here. 4. **Be concise.** One aggregated report. Issue messages one sentence; recommendations map to a concrete next edit (extract X, lift Y to context). ## Parse Arguments Arguments: $ARGUMENTS Optional: - `--path <dir>`: target directory (default: current working directory) - `--since <ref>`: target files changed since a git ref (`git diff --name-only <ref>...HEAD`) - `--all`: review all frontend component files in the repository - `--json`: emit the agent's aggregated JSON instead of prose (for CI) ## Steps ### 1. Determine target files Same auto-scope logic as `/code-review`: review uncommitted changes if any exist, else the full repository; honor `--since`, `--path`, and `--all`. Then keep only frontend component files — `.jsx`, `.tsx`, `.vue`, `.svelte`, Angular `*.component.ts` and their templates, and `.js`/`.ts` modules that render UI. Exclude `node_modules`, `dist`, `build`, and test files. If no frontend component files are in scope, emit `No frontend component files in scope — nothing to review.` and stop. ### 2. Dispatch the review agent Spawn `component-architecture-review` as a sub-agent in a single batch, passing the in-scope files as full files (`Context needs: full-file`). It returns its standard JSON (`status`/`issues`/`summary`). If it skips, report that and stop. ### 3. Aggregate and rank Take the agent's issues. Rank: semantic UI duplication and prop-drilling that force multi-site edits first (`error`), then high-value extraction and granularity fixes (`warning`), then composition and API-consistency cleanups (`suggestion`). Group by file. Drop any finding that belongs to another agent per constraint 3. ### 4. Report If `--json`, emit the aggregated JSON object and stop. Otherwise produce one report (chat for a small target; `.dev-team-reports/frontend-architecture-<date>.md` for a module): ```markdown ## Frontend Architecture Review — <target> **Health**: <pass|attention|critical> **Component files**: N **Findings**: N ### Findings (by severity) | File:line | Issue | Severity | Suggested fix | ### Extraction & composition recommendations <reusable components to extract, prop-drilling to lift, APIs to align, missing or stale component-library `index` barrel to add — each with the concrete next edit> ### Next steps - Mechanical fixes (extract component, call existing primitive) → /apply-fixes - Larger refactors (lift state, recompose) → /plan or /build ``` Surface only what's actionable. If everything is clean, say so in one line.
View on GitHub