| name | prime-frontend |
| description | Primes the agent with focused understanding of the frontend portion of the codebase — components, routing, state management, and styling — without loading unrelated backend code. Use at the start of a session when the work is scoped to UI or client-side features. Optionally pulls external task context from Jira issues and Confluence pages first. |
Prime Frontend: Load Frontend Context
Objective
Build targeted understanding of the frontend codebase by analyzing its structure, components, and conventions. Loading only frontend context keeps the context window light on complex full-stack codebases. If external task references are provided, load them first so the analysis is anchored to the actual work.
Scope discipline: Limit all file reads to the frontend root and its dependencies. Do NOT load the entire codebase — on large repos this exhausts the context window before any useful work is done.
Process
Step 0: Load External Context
Run this step BEFORE the codebase analysis. It accepts optional arguments: [jira-issue-keys] [confluence-page-ids].
- Jira keys may be a single key (
PROJ-12) or comma-separated (PROJ-12,PROJ-13).
- Confluence page ids are numeric page ids.
If Jira issue keys are provided:
- Call
mcp__atlassian__getAccessibleAtlassianResources to obtain the cloudId.
- For each Jira key, call
mcp__atlassian__getJiraIssue with that cloudId, the issue key, and responseContentFormat: "markdown".
- Treat the returned issue summary, description, and acceptance criteria as the task context for everything that follows.
If Confluence page ids are provided:
- Call
mcp__atlassian__getConfluencePage for each page id with contentFormat: "markdown" (use the cloudId from above, fetching it via mcp__atlassian__getAccessibleAtlassianResources if it was not already retrieved).
- Treat the returned page content as supporting context (specs, design docs, requirements).
If no arguments are provided: Skip this step entirely and proceed to Step 1.
Briefly summarize any external context loaded before continuing — this frames the rest of the priming.
1. Locate the Frontend
List all tracked files to find the frontend root:
!git ls-files
Common frontend roots: frontend/, client/, web/, src/ (when project is frontend-only), app/ (Next.js). Identify the correct root before proceeding.
2. Read Frontend Documentation
- Read CLAUDE.md or similar global rules file (for project-wide conventions)
- Read any README inside the frontend root
- Read
.claude/references/frontend-component-best-practices.md if it exists — it contains project-specific component conventions
3. Identify Key Frontend Files
Based on the structure, read:
- Main entry point (
main.tsx, index.tsx, app/layout.tsx, pages/_app.tsx, etc.)
- Routing configuration (
router.tsx, routes.ts, app/ directory for Next.js)
- Global state setup (store, context providers)
- Shared component library root (
components/, ui/)
- Core configuration (
package.json, tsconfig.json, vite.config.ts, next.config.ts)
- One or two representative feature components to internalize the established patterns
Skip files outside the frontend root unless they define a shared type or contract the frontend depends on.
4. Understand Current Frontend State
Check recent frontend-relevant activity:
!git log -10 --oneline
!git status
Note any open changes in the frontend directory.
Output Report
Provide a concise summary covering:
External Task Context (if loaded)
- Jira issue(s): key, title, one-line goal, acceptance criteria
- Confluence page(s): title and what they specify
Frontend Overview
- Framework and major libraries (React, Vue, Next.js, Tailwind, etc.)
- Component patterns observed (atomic design, feature folders, etc.)
- State management approach
Directory Map
- Frontend root and key sub-directories with one-line purpose each
Conventions
- Naming conventions, file co-location rules
- Styling approach
- Testing framework and conventions observed
Current State
- Active branch, recent frontend changes
- Any immediate concerns (missing types, deprecated patterns, etc.)
Make this summary easy to scan - use bullet points and clear headers.
