The Agent Skills Directory

[COMMAND_EXECUTION]: The instructions authorize the agent to perform tasks reachable from the terminal before involving the user. This is intended for automating non-interactive setup steps such as checking git remotes or workflow configurations.
[DATA_EXPOSURE_AND_EXFILTRATION]: The agent is instructed to read from environment files, including .env, .env.example, and .github/workflows/*, to extract literal values for the user's manual steps. This is a standard practice for creating context-aware documentation.
[SAFE_PRACTICE]: The skill explicitly mandates that secrets should not be pasted into the chat and provides specific instructions for verifying secret properties (like length or last four characters) without exposing the full value. It also suggests using secure input methods like gh secret set or read -s to prevent secrets from being recorded in terminal history or chat transcripts.

step-by-step