prompt logo

prompt

Free

Adversarial AI verification for software changes

FreeFree tier
Inputs: textOutputs: text
Type
Open Source

About prompt

The Verification Specialist Agent prompt is a detailed system prompt designed for AI agents to perform rigorous verification of software changes. It instructs the agent to actively try to break implementations rather than simply confirm they work, covering frontend/UI, backend/API, CLI/script, infrastructure/config, and library/package verification. The prompt includes critical rules such as not modifying the project, checking available tools, and providing strategies for different types of changes. It is part of the ai-boost/awesome-prompts open-source collection.

Key Features

Defines two common verification failure modes: check-skipping and being lulled by the obvious 80%
Prohibits the agent from modifying project files or installing dependencies
Provides verification strategies for frontend/UI, backend/API, CLI/script, infrastructure/config, and library/package
Includes spot-check warning that any claimed command output will be re-executed
Supports use of browser automation MCP tools if available
Specifies that test scripts must be written to /tmp and cleaned up

Pros & Cons

Pros
  • Encourages thorough, adversarial testing rather than superficial confirmation
  • Provides specific strategies for different types of software changes
  • Includes explicit warnings against common verification pitfalls
  • Designed to be reusable across various projects and tasks
  • Enforces accountability with spot-check warnings
Cons
  • Requires an AI model capable of executing commands and browser automation
  • May be overly complex for simple changes that don't need adversarial testing
  • Does not include built-in test frameworks; relies on external tools and commands
  • Limited to the capabilities of the underlying AI agent (e.g., no actual GUI interaction without MCP)

Best For

Automated quality assurance of pull requests and code changesTesting frontend UIs via browser automation and resource loadingValidating backend API endpoints including error handling with malformed inputVerifying CLI tools with edge-case inputsChecking infrastructure configurations with dry-run commandsLibrary and package testing by building artifacts and running test suites

FAQ

What is the Verification Specialist Agent prompt?
It is a system prompt designed for AI agents to perform rigorous verification of software implementations by actively trying to break them, rather than simply confirming they work.
What types of verification strategies are included?
The prompt covers frontend/UI, backend/API, CLI/script, infrastructure/config, and library/package verification, with specific instructions for each.
Can the agent modify project files?
No, the agent is strictly prohibited from creating, modifying, or deleting any file inside the project directory. It may write short-lived test scripts to /tmp or $TMPDIR and must clean them up.