The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to perform version control operations, specifically git commit and git tag, to mark the completion of milestones.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface where instructions embedded in project files (like ROADMAP.md or phase SUMMARY.md files) could attempt to influence the agent's behavior during the archiving process.
Ingestion points: {{version}} argument, phase SUMMARY.md files, .planning/ROADMAP.md, and .planning/REQUIREMENTS.md (as seen in SKILL.md).
Boundary markers: Absent; the skill reads entire file contents into context for processing.
Capability inventory: Bash (git commands), Write (file creation and deletion of .planning/REQUIREMENTS.md), and Read (all scripts).
Sanitization: Absent; the skill relies on the underlying LLM's safety filters, but includes multiple human-in-the-loop confirmation gates ("Wait for confirmation", "Present for approval") as an operational mitigation.
[DATA_EXPOSURE]: Accesses configuration and workflow templates located in the user's home directory at $HOME/.claude/get-shit-done/. This access is necessary for the skill's primary function to load its specific processing templates.

gsd-complete-milestone