Visual Guide
Step-by-step screenshots for running audits in every tool. Same process everywhere: open your project, copy the prompt, paste it, get results.
Not sure which format to use? See the format guide below.
Choose Your Tool
Claude Code (CLI)
Terminal format- 1
Open your project in the terminal
Navigate to your project directory and start Claude Code. Your full codebase is now accessible.
~/projects/taco-tracker╭─ Claude Code v2.1.83 ─╮
Welcome back!
Opus 4.6 (1M context)
~/projects/taco-trackerTips for getting
started
✔ Run /init...
Recent activity
No recent activity❯ |8 shells · ►► accept edits on - 2
Copy the prompt from AuditBuffet
On AuditBuffet, go to the audit you want to run (or the Stack Scan). Make sure "Terminal" format is selected, then click "Copy Prompt."
auditbuffet.com/audits/securityAuditBuffetAuditsPacksDocsDashboard← Audit Catalogv1.2FreeSecurityOWASPSecurity AuditComprehensive security analysis covering OWASP Top 10, authentication, input validation, and dependency risks.What's Included24Total Checks3Delivery Formats4Categories4VersionsChangelogv1.2Hardened curl commands, added response validation - 3
Paste the prompt and press Enter
Paste the full prompt into the Claude Code chat. It reads your files, runs the checks, and submits the JSON results to AuditBuffet automatically.
~/projects/taco-tracker❯ [Pasted text #1 +1092 lines]Running security audit on taco-tracker...
⏺ Reading src/taco-map.tsx...
⏺ Reading src/salsa-engine.ts...
⏺ Checking OWASP Top 10...
✓ 18/24 checks passed
⚠ 4 warnings, 2 critical
⏺ Submitting via curl...
✓ Results submitted to auditbuffet.com8 shells · ►► accept edits on - 4
See your results on AuditBuffet
Your project and scores appear in your dashboard within seconds. No copy-pasting results — it's all automatic.
auditbuffet.com/dashboard/taco-trackerAuditBuffetAuditsPacksDocsDashboardtaco-tracker3 audits completed · Last run: just nowBBSecurity78%18/24 checks passedAAccessibility92%22/24 checks passedCPerformance64%14/22 checks passed
Understanding Prompt Formats
Every audit has up to three prompt formats. Pick the one that matches your tool — it determines how the audit reads your code and returns results.
Terminal
A single prompt that reads your codebase directly. The AI has full file access and submits results automatically via curl.
Best for:
Claude Code (CLI & Desktop), Cursor, VS Code + AI extensions, Windsurf, Aider, Replit Agent
Browser
The audit is split into smaller chunks you paste one at a time. Designed for tools with smaller context windows that already have your code loaded.
Best for:
Bolt, Lovable, v0, Base44
Chat
The AI asks you to paste config files and code snippets interactively. You copy the JSON output and submit it manually on AuditBuffet.
Best for:
ChatGPT, Claude.ai, Gemini, any general-purpose chat interface
Still stuck?
The process is the same for every tool: open your project, copy the prompt, paste it, done.