The Agent Skills Directory

[COMMAND_EXECUTION]: The skill leverages standard software development tools (e.g., tsc, npm, eslint, mypy, cargo clippy) to perform mandatory static analysis and build verification as part of the §3 Verification Gate.
[PROMPT_INJECTION]: The instructions include specific safety triggers (ESCALATE) that require explicit user approval for risky operations such as deleting files, dropping tables, or modifying security-sensitive code (e.g., auth logic).
[EXTERNAL_DOWNLOADS]: The skill is designed to retrieve current library documentation and release notes from trusted domains and well-known repositories (e.g., biomejs.dev, github.com/DietrichGebert/ponytail) to ensure API compatibility and security compliance.
[REMOTE_CODE_EXECUTION]: Long-running tasks such as CI/CD pipelines and external deployments are managed through the cli-jaw background task interface, which is scoped strictly to development-related workflows.
[DATA_EXFILTRATION]: While the skill processes user code and project structure, it lacks any instructions or mechanisms for unauthorized network transmission of sensitive data; operations are confined to local build, test, and documentation tasks.

dev