The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes Python's subprocess module to perform local version control operations using git and to coordinate its internal scripts. These operations are implemented using secure list-based arguments, preventing shell injection risks.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from repository files to construct its memory entries, creating a potential surface for indirect prompt injection.
Ingestion points: lore init scans repository metadata and documentation files; lore sync processes code changes via git diff output.
Boundary markers: The skill employs clear Markdown section headers (e.g., ## Lore (auto-managed)) and horizontal rules to isolate AI-generated content from manual user notes.
Capability inventory: The agent can write to the .lore/ directory and update platform-specific instruction files like CLAUDE.md. It executes git CLI tools to retrieve project history.
Sanitization: The workflow is designed around user oversight; every proposed modification to the memory bank or platform mirrors must be explicitly confirmed by the user before being applied.

lore