The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines patterns for integrating Claude Code via shell piping, which creates a vulnerability surface for indirect prompt injection.
Ingestion points: The cc:review task in examples/examples.md pipes git diff --staged into Claude, and cc:explain pipes file content via cat into a Claude session.
Boundary markers: No delimiters or "ignore instructions" warnings are present in the provided task definitions to distinguish data from instructions.
Capability inventory: The skill has extensive shell execution capabilities via the [tasks] section, file writing capabilities (e.g., creating .mise.local.toml), and tool management features.
Sanitization: No sanitization or validation of the piped external data is described.
[COMMAND_EXECUTION]: The skill is a task runner configuration that explicitly defines and executes arbitrary shell commands. While this is the intended primary purpose, it allows for significant interaction with the host system's shell environment through mise.toml task definitions and executable scripts in .mise/tasks/.
[PERSISTENCE]: The skill documentation encourages the installation of git hooks via hk and pre-commit (prek). These tools install executable scripts into the .git/hooks/ directory, which triggers automatic execution during standard git operations.
[EXTERNAL_DOWNLOADS]: The skill is designed to manage and install external tool binaries and plugins.
It downloads custom plugins from github.com/jdx/mise-env-fnox.
It fetches configuration and builtin definitions for the hk hook manager from github.com/jdx/hk release assets.
These downloads target the official repositories of the tool author and are consistent with the skill's stated purpose.

mise