The Agent Skills Directory

[DATA_EXFILTRATION]: The skill includes robust safeguards to prevent accidental leakage of sensitive information. The archiving logic in scripts/archive.py automatically excludes known secret files and directories, such as .env files, SSH keys, and AWS credentials, before any data is transmitted to the backend.
[COMMAND_EXECUTION]: Local command execution is limited to legitimate project tasks, such as using git to identify files to include in the migration while respecting project ignore rules. The use of subprocess is scoped to local execution of bundled or standard system tools and does not involve executing untrusted strings.
[EXTERNAL_DOWNLOADS]: The skill safely interacts with official GitHub and AWS services. It implements strict domain validation for its API endpoints and restricts redirect hosts during GitHub downloads to trusted domains, effectively preventing SSRF and unauthorized data redirection.
[CREDENTIALS_SAFE]: The skill avoids hardcoded secrets, instead using a secure OIDC PKCE authentication flow. Authentication tokens are stored locally with restricted filesystem permissions (mode 0600), adhering to industry best practices for secure token management in CLI tools.

launch-with-aws