How to use the Business Logic Attack Testing agent skill

What is an agent skill, and what does this one do?

An agent skill is a folder of instructions that an AI agent loads when a task matches it. This one is named offensive-business-logic and is distributed as a single file, SKILL.md, inside the SnailSploit/Claude-Red repository under Skills/web/.

Once installed, the agent maps a multi-step flow such as checkout or account setup as a state machine: the states, the allowed transitions between them, and the side effects of each. It then checks each transition against four axes (state, identity, input and frequency) and suggests specific probes, such as skipping a payment step, sending negative quantities, swapping currencies, stacking coupons or replaying a refund.

• Covers workflow bypass, price, quantity and currency manipulation, coupon logic, refund and payout abuse, role and tenant boundaries, and race conditions on logic checks.
• Examples in the skill are curl and HTTP request snippets.
• The skill is a methodology. The agent still has to reason about each target.

How do you install the skill?

Download the ZIP for this skill from this page. Then choose one of two routes.

For Claude Code, unzip it into your personal skills folder ~/.claude/skills/, or into a project's .claude/skills/ folder if you want it limited to one project.

In the Claude apps, open Settings, go to Capabilities, and upload the skill under Skills.

• Personal install: ~/.claude/skills/ (available across your projects).
• Project install: .claude/skills/ inside the project folder.
• Claude apps: Settings → Capabilities → Skills, then upload.

How do you trigger the skill with a natural request?

You do not need a special command. Describe the target and the kind of flaw you are after, and mention that the work is authorised. The skill's description says it is meant for applications after surface-level OWASP Top 10 coverage, or for transactional, marketplace, fintech, e-commerce and SaaS apps where logic flaws have direct financial impact.

Give the agent the flow to examine and the API or UI it can see. The more concrete the flow, the more specific the probes.

• Example: 'We have written authorisation to test our checkout. Map the states from cart to confirmation and suggest logic probes for each transition.'
• Example: 'Review this refund endpoint for ways to refund more than was paid.'
• Example: 'Look at our coupon API and list the ways a discount could stack or persist after it should have expired.'

What does the skill need, and what are its safety limits?

The listing states that the skill requires nothing beyond itself, runs no scripts and does not use the network. It is a set of instructions, so any requests the agent suggests are sent by you or by tools you run, against systems you control or have permission to test.

The techniques are attacks against live systems. Use them only with written permission. Testing an application you are not authorised to test may be illegal.

• Requires: nothing listed.
• Runs scripts: no.
• Uses network: no.
• Only test systems you own or have written permission to test.
• Keep test accounts and test payment data separate from real customer data.

Tips for getting useful results

Single-axis flaws are often low severity; the skill's own method says that chains are where the critical findings come from. Ask the agent to combine findings and to estimate financial impact per finding, such as loss per attack multiplied by scale.

Start from a state machine rather than a list of endpoints. Ask the agent to name the preconditions for each transition before it proposes any request.

• Give it the full flow, not one endpoint.
• Ask for chains of two or more flaws, not isolated ones.
• Ask for loss per attack and how it scales, so findings can be prioritised.