The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands for version control and CI/CD operations. This includes standard git commands, GitHub (gh) and GitLab (glab) CLI tools, and local Python validation scripts (medic.py, council.py). These commands are used for their intended purpose within the scope of code management.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data during the history review and Pull/Merge Request listing phases.
Ingestion points: The agent reads data from git log, gh pr list, and glab mr list, which contain commit messages and PR titles/descriptions authored by other users (SKILL.md).
Boundary markers: The skill implements strict Human-in-the-Loop (HITL) checkpoints, requiring the agent to "DỪNG chờ duyệt" (Stop for approval) and display a checklist before performing any repository-altering actions.
Capability inventory: Includes local file writing (RELEASE-vX.Y.Z.md), remote repository interactions (push, tag, merge), and execution of local CI/CD scripts.
Sanitization: Security is maintained by requiring explicit user confirmation for all side-effect commands, allowing the user to inspect any content derived from external inputs before it is executed or published.

ship