The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon project-specific configuration files (such as manifests and dependency lists) which may contain untrusted data. A maliciously crafted project could use these files to trigger unintended actions when the agent activates the environment.
Ingestion points: The skill reads content from .flox/env/manifest.toml, package.json, and requirements.txt within the project root.
Boundary markers: None. The instructions do not specify the use of delimiters or 'ignore' instructions for the content parsed from these project files.
Capability inventory: The skill enables shell command execution via on-activate hooks and profile blocks, package management via flox install, and the management of background services (e.g., PostgreSQL, Redis).
Sanitization: No sanitization or validation of the hook scripts or manifest content is performed; the shell commands are executed directly as defined in the project configuration.
[COMMAND_EXECUTION]: The skill utilizes shell-based hooks (on-activate) to automate environment setup tasks, such as running package managers (npm install, uv pip install) and initializing databases (initdb). While these are core features of the tool, they allow for the execution of arbitrary shell commands based on the contents of the local .flox/ directory.

flox-environments