More screenshots
Works with the AI Agents you already use
Red-Team Thinking: Attack Your Own System First
Plan and run an authorized adversarial review of your own code, auth, or API — model the attacker, map the surface, and report only findings with a reproducible exploit path.
$5
Red-Team Thinking: Attack Your Own System First
Example session with this skill installed
We are about to ship a new login flow with password reset and session cookies. Red-team it before launch and tell us what an attacker could actually break, with proof.
- Read your context and instructions
- Compiled the red-team thinking
Scoped the review first: the login form, the password-reset flow, the session cookie, and the account-lockout logic were in bounds; the production database and other tenants were explicitly out of bounds. Modelled the attacker as a single unauthenticated outsider with a browser and a script, then mapped the surface: three endpoints, two token types, one cookie. Found two real issues with reproducible paths. First, the reset token was accepted twice — requesting a reset, using the link, then reusing the same link a second time still changed the password, because the token was only marked used after a delay. Second, the lockout counted attempts per username only, so an attacker could spray one password across many accounts without ever tripping it. Two suspected issues did not reproduce and were reported as tested-and-clear, with the exact requests run. The session cookie flags, the reset-token entropy, and the rate limit on the login endpoint all held. Each confirmed finding shipped with the exact request sequence, the observed response, and a verified fix: single-use tokens invalidated at first use, and lockout keyed on the source as well as the account. The review closed with the scope, the evidence log, and the two fixes re-tested — both now fail to reproduce, which is the only result that counts as done.
Connects securely to your tools. The creator never sees your data.
About this skill
Most security reviews fail the same way: they list what could theoretically go wrong and stop there. This skill plans and runs an authorized adversarial review of a system you control — code, auth, or an API — by modelling the attacker first, mapping the attack surface second, and testing third. Every finding it reports carries a reproducible exploit path and a verified mitigation, so the review ends with fixes, not with a pile of warnings nobody can act on. It also covers scope discipline: what is in bounds, what is out of bounds, how to record evidence without causing harm, and how to stop the moment the work leaves authorized ground.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Security.
It covers systems you control or are authorized to test; it never authorizes testing someone else's system.
A finding counts only with a reproducible path — anything that cannot be reproduced is reported as unconfirmed, not as a vulnerability.
It is a review method, not a scanner: it does not replace automated tooling or a licensed penetration test where one is required.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 3 days ago
Permissions
No special permissions declared or detected