The Agent Skills Directory

[COMMAND_EXECUTION]: Executes diagnostic shell scripts (check-health.sh and check-health.ps1) and runs system commands for Git, GitHub CLI, and package managers (NPM, Homebrew) to manage environment state and tool installation.
[EXTERNAL_DOWNLOADS]: Downloads and installs various development tools and agent extensions from public registries and GitHub repositories. Sources include well-known entities like Microsoft and Vercel Labs.
[PROMPT_INJECTION]: The skill ingests untrusted configuration data from repository-specific GitHub issues. This creates an indirect prompt injection surface where external data influences tool selection and workflow behavior.
Ingestion points: gh issue view retrieves configuration from the body of a GitHub issue labeled tunan:config (referenced in SKILL.md and references/config-issue-storage.md).
Boundary markers: Data is expected within markdown YAML code blocks.
Capability inventory: Shell execution (tool installs), file deletion (tunan.local.md), and GitHub API writes across scripts and instructions.
Sanitization: No explicit validation or filtering of the ingested YAML configuration is observed.
[DYNAMIC_CONTEXT_INJECTION]: Employs dynamic context injection (!command syntax) in SKILL.md to execute echo commands at load time for platform identification and environment configuration.

setup