The Agent Skills Directory

[SAFE]: The skill acts as a high-quality knowledge base and procedural guide for AI agents to perform security audits. It uses structured personas and rigorous methodology to identify vulnerabilities in blockchain code.
[NO_CODE]: The skill consists almost entirely of markdown documentation, vulnerability checklists, and process guides. It does not ship with standalone executable scripts or binaries.
[COMMAND_EXECUTION]: Use of local shell commands (like grep, find, and cat) is restricted to the detection of the skill's own installation directory and the scanning of the project files provided for audit. These operations are standard for security analysis tools and are performed within the local execution environment.
[DATA_EXFILTRATION]: No network-based operations or attempts to transmit data to external servers were identified. The framework instructs the agent to save all audit outputs to a local .context/outputs/ directory for user review.
[PROMPT_INJECTION]: The instructional language is designed to set a professional persona ('senior smart contract security auditor') and enforce strict analytical rigor (e.g., 'MANDATORY DEBUG LOGGING'). It does not attempt to bypass platform safety filters or override the base identity of the agent.

smart-contract-audit