The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local CLI tools including git to inspect repository state, gh to create issues and labels on GitHub, and node to run a bundled wave-sorting script (wave-sort.mjs) for generating plan artifacts.
[EXTERNAL_DOWNLOADS]: The generated dag.html file includes stylesheet references to well-known Google Fonts services (fonts.googleapis.com and fonts.gstatic.com) for rendering the interactive graph.
[PROMPT_INJECTION]: The skill processes project documentation and repository history to draft sprint plans. While this represents a surface for indirect prompt injection from potentially untrusted project files, the risk is addressed by a mandatory human approval step (Step 6) that requires the user to review the entire plan before any GitHub issues are created.
Ingestion points: Step 1 reads repository files (CLAUDE.md, README) and merged PR history to infer next tasks.
Boundary markers: No explicit boundary markers are specified for the ingested content in the drafting instructions.
Capability inventory: The skill can create GitHub issues and labels via the gh tool and write artifacts to the local .claude/sprints/ directory.
Sanitization: The wave-sorting script provides character escaping for graphical output, but the issue creation process relies on the AI to summarize content safely. The human-in-the-loop requirement provides a critical verification checkpoint.

sprint-plan