The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill's CI/CD workflow generation template includes a command to fetch and execute the Railway CLI installer (curl -fsSL railway.com/install.sh | bash). This practice is documented as it targets the official domain of a well-known service.
[COMMAND_EXECUTION]: The skill performs extensive shell command execution for project state inspection, Git lifecycle management (git add, git commit, git revert), and interaction with various cloud provider CLIs (vercel, railway, wrangler, supabase, gh). These operations are necessary for the skill's functionality and are performed with clear user-facing logging.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection as it reads and parses metadata from project-level configuration files such as package.json, Cargo.toml, and railway.toml to determine the project's scenario and configuration.
Ingestion points: The skill reads various configuration files (package.json, turbo.json, railway.toml, wrangler.toml, etc.) to classify the project.
Boundary markers: Content generated in AGENTS.md is strictly delimited by a ## Make targets header and a unique lead paragraph to distinguish it from user-authored prose.
Capability inventory: The skill can write files (Makefiles, documentation, CI/CD YAML), execute shell commands and external CLI tools, and perform network requests via deployment tools.
Sanitization: Data extraction from project files is performed using standard tools like jq and grep rather than direct interpolation into instructions.

init-makefiles