The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The skill utilizes dynamic context injection in SKILL.md to execute probe-bun-mcp.sh at load time. This script performs benign environment checks (using git and jq) to determine if the Bun documentation MCP server is configured in the current project, allowing the agent to provide tailored instructions for documentation lookups.
[EXTERNAL_DOWNLOADS]: The gh-release example demonstrates fetching release metadata from the GitHub API and downloading assets. The implementation includes retry logic and status checks, and the associated documentation in references/bun-api.md explicitly warns about validating untrusted archive contents during extraction.
[COMMAND_EXECUTION]: The skill provides detailed patterns for using Bun Shell ($) and Bun.spawn(). It includes specific security sections warning against bash -c with user-controlled strings and explaining how to prevent argument injection by validating inputs against target command semantics.
[DATA_EXPOSURE_&_EXFILTRATION]: The env-sync example provides a robust pattern for managing sensitive .env files. It uses the Web Crypto API (AES-GCM with PBKDF2 key derivation) for encryption and decryption, and it implements interactive password prompts rather than hardcoding or logging secrets.

bun-cc