The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructions advise the agent to provide the user with a command (curl https://mise.run | sh) to install the mise utility. Although this is the tool's official installation method, piped shell execution from remote URLs is a high-risk pattern.
[DYNAMIC_CONTEXT_INJECTION]: The skill uses the !command`` syntax to execute shell commands like pwd and find when the skill is loaded. These commands are used to identify existing configuration files (e.g., mise.toml, .tool-versions) and project structure for context gathering.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes various project-controlled configuration files and has the capability to execute shell commands (Bash) and modify the filesystem. A malicious configuration file or Makefile in a project directory could potentially influence the agent's behavior during migration or audit tasks.
Ingestion points: Reads project configuration files including mise.toml, Makefile, Brewfile, and legacy version files like .tool-versions.
Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are present for the ingested file content.
Capability inventory: Uses Bash, Write, Edit, Read, and Grep tools.
Sanitization: No specific sanitization or validation of the ingested file content is mentioned before it is processed or used to influence agent decisions.

configure-mise