The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on extensive shell command execution using the gh (GitHub) and git command-line interfaces. These commands are used to manage project resources, such as creating milestones, labels, issues, and project boards, as well as managing code via git worktrees, branches, and pull requests. Additionally, in Phase 5, the skill generates a task-specific SKILL.md file at the project level which contains executable instructions and checklists for the agent to follow in subsequent sessions.
[PROMPT_INJECTION]: The workflow's planning and skill generation phases are informed by the analysis of external project data. This represents a potential indirect injection surface where malicious patterns within an analyzed codebase could attempt to influence the logic of the generated task breakdown or the project-specific sub-skill.
[SAFE]: The skill performs deep project analysis that reads the current codebase to generate architecture overviews and module inventories. This technical metadata is stored in local documentation and GitHub Issues as part of the intended workflow. All network operations are directed to the well-known and trusted domain github.com using the official GitHub CLI tool.

spec-driven-develop