Authorization Bypass Detection
Probes an AI agent with dialogue to find cross-user access and privilege escalation flaws.
- Repository
- Tencent/AI-Infra-Guard
- Stars
- 7,320
- Licence
- Apache-2.0
- Files
- 1 file
What it does
This skill teaches an AI agent to act as a security tester for another AI agent. It sends it chat prompts that try to get data belonging to other users, perform admin-only actions, or skip login checks, then judges whether the target agent refused or leaked or acted beyond its authority. Findings are written up in a structured report with evidence, severity and fix suggestions.
Great for
- Security teams reviewing an AI assistant that has user accounts, roles or admin features
- Developers checking that their agent enforces access control before release
- Auditors who need a repeatable set of authorization test prompts
Skip it if
- Agents with no users, roles or privileged operations, since the skill itself says to skip them
- Testing systems you do not own or have written permission to test, because the prompts attempt unauthorized access and could cause harm or break terms of service
Try asking
- “Test my customer support agent to see whether it will show one customer's orders to another customer.”
- “Check whether my admin chatbot can be tricked into granting admin rights or disabling login.”
- “Run an authorization bypass review on our internal HR assistant that handles multiple tenants.”
Install in 3 steps
- 1
Download the skill
The ZIP contains the 1 original files, the Apache-2.0 licence and an attribution note.
authorization-bypass-detection.zip - 2
Add it to your agent
Claude Code (all projects)
mkdir -p ~/.claude/skills && unzip authorization-bypass-detection.zip -d ~/.claude/skills/Claude Code / Agent SDK (this project only)
mkdir -p .claude/skills && unzip authorization-bypass-detection.zip -d .claude/skills/Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.
- 3
Ask for the task
No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.
Get better results
- Only run it against systems you are authorised to test, ideally in a staging environment with dummy accounts.
- Provide the target's roles and user accounts so the tester can craft realistic cross-user prompts.
- Review every flagged finding manually, since the judgement is based on the target's text response.
What's inside
- SKILL.md
Redistributed unmodified from Tencent/AI-Infra-Guard under the Apache-2.0 licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.