Prompt di correzione AI

Come funziona

Al clic si combinano due pezzi di dati:

• The finding — the issue summary, affected surface, remediation guidance, and safe evidence needed to help your coding agent fix it.
• Your app context — FixVibe uses scan context when available to choose a framework-aware remediation shape, and falls back to a generic recipe when it cannot infer enough context.

Fix prompts are rendered server-side from FixVibe remediation guidance. They are designed for copy-paste use in Cursor, Claude Desktop, Copilot, or another coding agent without exposing the internal prompt registry in the browser.

Com'è fatto il prompt

Fix the "Reflected XSS in /search?q=" vulnerability at /search.

Issue: Query parameter q is rendered into the response body without
escaping; an attacker can inject <script> via crafted URLs.

Codebase context: Next.js.

Recommended fix:
In Next.js, render user-supplied values through JSX ({value}) so React's
automatic escaping kicks in. For server components rendering rich HTML,
sanitize with DOMPurify (server-side via JSDOM) before output.

Constraints:
- Don't break existing tests; run the test suite after the change.
- Match the codebase's existing style and lint config.
- Add a brief comment explaining the security reasoning only where the
  fix would otherwise look arbitrary.
- If the fix needs a new dependency, install it via the project's
  package manager (npm / pnpm / pip / bundle / composer).

Reference: CWE-79 — see https://cwe.mitre.org/data/definitions/79.html

Framework supportati

Mostriamo snippet specifici per framework per:

• Next.js, React, Vue, Nuxt, Svelte (frontend)
• Express, Fastify (backend Node.js)
• Django, Flask (Python)
• Ruby on Rails
• Laravel (PHP)
• ASP.NET Core fallback guidance

Framework context is best-effort. If FixVibe cannot infer enough safely from the scan, the prompt asks your coding agent to inspect the repository before applying the fix.

Usalo dal tuo agente AI

Se hai collegato il server MCP, lo stesso prompt è esposto come comando slash. Da Claude Desktop:

/fixvibe-fix finding_id=550e8400-e29b-41d4-a716-446655440000

The server looks up the finding, applies available scan context, renders the remediation prompt, and injects it into your conversation as the user message. No third-party LLM API call is made by FixVibe for this templated prompt.

Perché non chiamiamo Claude a ogni clic

Al lancio abbiamo considerato di chiamare la Anthropic API a ogni clic per rifinire il prompt con il contesto del codebase. Non l'abbiamo fatto, perché:

• L'agente in cui l'utente incolla il prompt ha già il contesto del codebase: sta usando Cursor / Claude Desktop con il repo aperto.
• Server-side templating covers the common remediation paths without any per-click model call.
• Un opt-in “Rifinisci con AI per il mio codebase” potrebbe chiamare l'API più avanti se gli utenti lo vogliono. Oggi no.