The Agent Skills Directory

[COMMAND_EXECUTION]: The skill employs Python scripts to manage agent files by scanning directories, moving files, and reading content.
Evidence: scripts/init_registry.py uses shutil.move to transfer agent files from standard Claude Code locations (~/.claude/agents/) to the skill's internal directory for management.
[EXTERNAL_DOWNLOADS]: The installation script downloads a standard library for interactive command-line interfaces.
Evidence: install.sh runs pip3 install questionary to facilitate the interactive selection of agents during the migration process.
[PROMPT_INJECTION]: The skill serves as a management layer for external agent instruction files, creating a surface for indirect prompt injection.
Ingestion points: scripts/init_registry.py and scripts/rebuild_registry.py read and parse Markdown files from external agent directories.
Boundary markers: Absent; the skill does not use specific delimiters or instructions to prevent the agent from obeying instructions embedded in the loaded files.
Capability inventory: The skill can read and move files, and it returns full file contents to the agent via scripts/get_agent.py.
Sanitization: No sanitization or content validation is performed on the ingested agent files.

agent-registry