The Agent Skills Directory

[SAFE]: The skill documentation includes explicit safety directives that prohibit deceptive humanization techniques such as inserting invisible characters, zero-width spaces, or intentional typos. These instructions prevent the agent from using common obfuscation methods to bypass detection.
[SAFE]: The included utility script scripts/validate_receipt.py is designed for local integrity verification using SHA-256 hashing. It is implemented with only Python standard libraries and contains robust security controls, including path resolution and validation to prevent directory traversal.
[SAFE]: Data processing instructions require the agent to maintain a claim inventory and perform factual verification against official documentation. This structured workflow serves as a mitigation against indirect prompt injection by ensuring that untrusted input is analyzed for substantive claims before being transformed.
[SAFE]: The skill maintains a restricted execution environment. It explicitly denies authorization for network operations, paid service subscriptions, or automated browser control, ensuring that all external interactions remain under direct user supervision.

anti-detection-writing