The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to download Python dependencies via pip and fetches external plugin tools from a remote repository (github.com/richfrem/agent-plugins-skills) using the uvx tool.
[REMOTE_CODE_EXECUTION]: The skill utilizes uvx to execute code directly from a remote Git repository to install or run plugins. It also executes a local initialization script (scripts/init.py) to scaffold the environment profile.
[COMMAND_EXECUTION]: The agent runs several shell commands for environment setup, including find for directory discovery, pip-tools for dependency management, and python for running the setup and verification scripts.
[INDIRECT_PROMPT_INJECTION]: The skill ingests directory names discovered on the filesystem and user-provided paths to build a configuration manifest, which creates a surface for potential indirect instructions.
Ingestion points: Output from the find . -maxdepth 1 -type d command and manual user input for source folders provided during the setup process.
Boundary markers: The skill requires the agent to present the finalized manifest to the user for explicit approval before writing the file to the filesystem.
Capability inventory: The skill has access to the Bash tool for package installation and execution of setup scripts, and the Write tool for persistent configuration storage in the .agent/learning/ directory.
Sanitization: The instructions do not specify any validation, filtering, or escaping for the discovered folder names or user-supplied paths before they are embedded in the JSON manifest.

vector-db-init