The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides extensive templates and documentation for executing shell commands via mise tasks. This includes operations for database migrations, build pipelines, and CI/CD workflows. While powerful, these are standard developer tools and the instructions emphasize safe patterns such as using read_file instead of exec to avoid process overhead.
[REMOTE_CODE_EXECUTION]: In references/advanced.md, the skill documents the experimental mise feature include, which allows importing task definitions from remote URLs (e.g., https://example.com/tasks.toml). This capability allows the execution of remote code if the source is not trusted.
[EXTERNAL_DOWNLOADS]: The skill recommends several third-party developer tools and libraries across different ecosystems, providing instructions to install them via pip, npm, and bun. This includes tools like Pants for build orchestration and semantic-release for versioning.
[INDIRECT_PROMPT_INJECTION]: The references/bootstrap-monorepo.md file contains a meta-prompt designed to guide an AI agent through an autonomous project scaffolding process. This involves executing a long sequence of commands based on generated content. The documentation includes a constraint instructing the agent that 'The human will not touch any code', which increases autonomy and should be reviewed by users before execution.
[SAFE]: The skill actively promotes security best practices, such as the use of redact = true for environment variables and reading secrets from local files rather than hardcoding them or using dangerous shell execution patterns.

mise-tasks