The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides patterns for wrapping CLI tools like git and gh. It explicitly recommends using an injectable Runner interface to avoid direct shell execution in testable logic and suggests using structured JSON output from tools rather than scraping human-readable stderr, which is a secure practice.- [EXTERNAL_DOWNLOADS]: References standard GitHub Actions such as actions/setup-go and golangci/golangci-lint-action. These are well-known, official tools for Go development environments.- [PROMPT_INJECTION]: Addresses the processing of external data (API responses, changelog excerpts, and LLM context) that could potentially contain malicious instructions. The skill provides a specific mitigation helper to truncate text on UTF-8 rune boundaries to prevent data corruption and recommends validating untrusted bytes at the ingestion boundary.- [DATA_EXFILTRATION]: Provides best practices for outbound HTTP requests, including the necessity of setting explicit timeouts and User-Agent headers to ensure process stability and proper identification. It does not contain any patterns associated with credential theft or unauthorized data transmission.

building-go-projects