The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Python subprocess module in scripts/cdp.py to launch local web browsers for automated navigation and in scripts/crypto.py to manage encryption keys and perform file encryption/decryption using the age utility. Both implementations utilize the list-based argument format which effectively prevents shell injection vulnerabilities.
[EXTERNAL_DOWNLOADS]: The script scripts/badbool.py fetches the 'Big-Ass Data Broker Opt-Out List' from a public GitHub repository managed by a known privacy researcher. Additionally, scripts/registry.py fetches the official Data Broker Registry from the California CPPA government website. These downloads are central to the skill's data-broker coverage and originate from well-known/trusted sources.
[DATA_EXFILTRATION]: As part of its primary function, the skill sends emails containing personal identifiable information (PII) to data brokers to request data removal. The scripts/emailer.py module includes a critical security control that locks the recipient address to only those officially declared in the curated broker database, preventing the agent from being manipulated into sending PII to arbitrary external addresses.
[CREDENTIALS_UNSAFE]: The skill requires EMAIL_PASSWORD and BROWSERBASE_API_KEY environment variables to enable automated email sending and cloud-based browser scanning. The skill follows best practices by instructing users to store these in environment variables or a .env file rather than hardcoding them.
[DATA_EXPOSURE]: Personal information is stored in dossiers under the user's home directory. The skill enforces restrictive file permissions (0600) and provides an optional at-rest encryption feature using the age binary to protect sensitive data from unauthorized local access.

unbroker