MCP Server

Agent Skill

Install the fix-feedback skill to give your AI agent a one-command workflow for processing and fixing client feedback.

The fix-feedback skill teaches your AI coding agent the full Faster Fixes workflow: fetch pending feedback, locate the relevant code, apply fixes, and mark items as resolved.

It works alongside the MCP server — the MCP provides the tools, the skill provides the workflow.

Install

The skill is distributed via skills.sh and works with Claude Code, Cursor, Windsurf, Cline, and 40+ other agents.

npx skills add manucoffin/faster-fixes/packages/skills/fix-feedback

What it does

When you say something like "fix feedback" or "check for new client reports", the skill guides your agent through:

  1. Fetch — Calls list_feedbacks with status: "new" to get pending items
  2. Mark in progress — Updates each item to in_progress so your team knows it's being handled
  3. Locate and fix — Uses the component path, CSS selector, and screenshot to find and fix the code
  4. Mark resolved — Updates the status to resolved after the fix is applied
  5. Summarize — Reports what was fixed and what was skipped

The skill also handles read-only requests. If you just want to see what's been reported without fixing anything, it fetches and presents the feedback without modifying code or statuses.

When to skip

The skill automatically skips feedback it can't resolve inline:

  • Vague comments with insufficient context
  • Large feature requests that require design decisions
  • Changes that need external dependencies not in the project

Skipped items are flagged in the summary so you can triage them manually.

Prerequisites

The MCP server must be configured before the skill can work. The skill calls the MCP tools under the hood — without them, it has nothing to fetch from.