The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill uses .flox/env/manifest.toml files which contain on-activate hooks and profile shell code snippets that are executed automatically whenever the environment is activated.
[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and act upon configuration from untrusted project-level files, creating a potential attack surface for indirect prompt injection.
Ingestion points: Reads and modifies .flox/env/manifest.toml, package.json, and requirements.txt (SKILL.md).
Boundary markers: None present; the agent treats the manifest content as trusted instructions for environment setup.
Capability inventory: Executes shell commands via hooks and profile definitions, performs package installations (flox install, npm install, pip install), and manages background services (SKILL.md).
Sanitization: No escaping or validation is specified for content interpolated from these files into shell contexts.
[COMMAND_EXECUTION]: The skill provides numerous patterns for executing shell commands and package managers (npm, pip, uv, cargo, gcc) within the environment lifecycle.

flox-environments