The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell commands to manage milestone file history and project state.
Evidence: The skill uses mv to archive existing milestones.json files with a timestamp suffix and uses git add and git commit to finalize changes in the repository.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and analyzes potentially untrusted external content to generate its output.
Ingestion points: User-provided document paths ($ARGUMENTS), project-specific CLAUDE.md, documentation in the docs/ directory, and the project's source code discovered via glob and grep operations.
Boundary markers: Absent; the skill does not use specific delimiters or instructions to isolate the data being analyzed from its primary operational instructions.
Capability inventory: Local command execution (mv, git), file system read access, and the ability to generate structured milestone data.
Sanitization: Absent; the skill directly processes the text from project files to determine the goals and verification steps for new milestones.

sefirot-milestone